Questions to Ask Before You Buy an AI Phone Answering System

Before buying an AI phone answering system, ask what it will do on a call, how and when it hands over to a person, where call data and recordings are stored, what it connects to, how you can test it with real calls, what happens past your allowance, and who fixes it when something goes wrong.

How to use this checklist

Most providers will tell you their system is easy and accurate. That tells you nothing. The questions below are designed to get specific answers you can compare between providers, and to show you quickly when someone is vague.

Take the list into every sales call, write down the answers, and ask for the important ones in writing. A provider that is happy to put a clear answer in an email or a contract is usually one you can work with. If you are still learning how the technology behaves, read what an AI voice assistant is and how it answers a call first.

What will it actually do on a call?

Start with the job, not the technology. Different systems do very different things: some only take a message, some answer common questions, some book appointments, some qualify enquiries. Be clear on which of these you are buying.

Our guide on what an AI voice assistant can and cannot do covers the realistic limits, so you can tell a fair answer from an overpromise.

Which tasks will it complete on its own, and which will it only pass on?

What does it do when a caller asks something outside its approved information? Ideal answer: it says it will pass the question on, rather than guessing.

Does it say it is an automated assistant, or does it pretend to be a person?

Which languages and accents has it been tested with, and how would I check that for my own callers?

What happens when a caller interrupts, goes quiet, or calls from a poor line?

How does handover and escalation work?

Handover is where small businesses get hurt, because a caller who needs a human and cannot reach one is worse off than with a plain voicemail.

Which situations trigger a handover (an urgent issue, an upset caller, a request for a person, a topic you have excluded)?

Where does the call go: a live transfer to a phone, a callback request, a text or email alert, or a message only?

What happens out of hours, or if nobody picks up the transfer?

Does the person receiving the call or message get a summary of what was already said, so the caller does not repeat themselves?

Can I change the handover rules myself or only through you, and how quickly?

What happens to your data and recordings?

A phone assistant handles names, numbers and often sensitive details, so treat this as a core buying question rather than a formality. The exact legal position depends on your country and sector, so these are the questions to put to the provider, not legal advice.

Recording and disclosure rules differ between countries. Our guide to AI call answering and recording rules by country is a starting point, and a qualified adviser should confirm the position for your business.

Where are recordings and transcripts stored, and in which country or region?

Who can access them: your staff only, the provider's staff, or other companies that supply parts of the service?

How long are they kept by default, and can you set a shorter period or have them deleted on request?

Does the provider use your calls to train or improve its own models, and can you opt out?

Ask for their data protection terms (for example a data processing agreement and a list of sub-processors) before signing, and read them.

How do they handle callers who object to being recorded, and who tells callers that a recording is happening?

What does it connect to?

An assistant that cannot see your diary or write to your records just creates more admin. Ask about the specific tools you use rather than accepting a general statement that it integrates with everything.

Calendar: does it read live availability and write bookings, or does it only send you a request to action by hand?

CRM or job software: which fields does it fill, and what happens when the caller is already in your system?

Phone system: can it work with your current number and provider, or must you port or change numbers? How are calls forwarded to it?

What happens to a booking if the connected tool is down?

Who sets up and maintains each connection, and is that included in the price?

How can you test it before launch?

Demos are polished. Real calls are not. The most useful thing you can do is test with your own material before you commit.

If a provider will only let you hear a scripted demo, treat that as an answer in itself.

Ask to hear real, unedited recordings of calls handled for other customers (with their permission and identifying details removed), including ones that went badly.

Write your own test scripts: your ten most common questions, a few awkward ones, a request it should refuse, and an urgent case that must reach a person. Call it yourself and have colleagues and friends with different accents do the same.

Ask whether there is a pilot or trial period, and what it costs.

Ask who reviews early calls after launch and how mistakes get corrected.

What does pricing and the contract really include?

Pricing varies widely between providers and is easy to compare badly, so ask for the whole picture in writing rather than a headline figure.

What exactly does the allowance cover (minutes, calls, or conversations), and what happens once you go past it? Is it a higher rate, a hard stop, or an automatic upgrade?

Are there setup, onboarding or integration fees, and what do they cover?

Are changes to scripts and rules charged separately?

What is the minimum term, and how do you cancel? Is notice required, and does the contract renew automatically?

If you cancel, can you export your call data and configuration, and when is your data deleted?

Who supports it after launch?

An assistant is not finished at launch. Your prices, opening hours and policies change, and callers will say things nobody planned for.

Who do I contact when something goes wrong, and what response should I expect?

How are changes made: do I edit them in a dashboard, or do I send a request to you? How long does that take?

Who reviews call performance over time and tells me what to fix?

What is the plan if the underlying speech or language service has an outage?

Which answers are red flags?

Promises of perfect accuracy or that it will never need a human.

No clear answer about where recordings are stored or who can access them.

Reluctance to share their data protection terms.

Only scripted demos, with no unedited calls and no trial.

A vague handover story, such as just saying it escalates when needed.

An allowance or overage rule that is hard to explain, or fees that only appear in the contract.

Long lock-in with no clear exit or data export.

No named person responsible for fixing problems after launch.

Erudience services related to this guide

Use cases related to this guide

Industries this guide applies to

Terms used in this guide

Frequently asked questions

What is the most important question to ask an AI phone answering provider?

Ask what happens when the assistant cannot help. A good provider explains clearly how it hands the caller to a person or records a message, and who gets alerted. Handover rules affect your callers more than any other feature.

Should I ask to hear real calls before signing?

Yes. Ask for unedited recordings of real calls handled for other customers, with their permission and personal details removed, and include some that went badly. Then test it yourself with your own scripts, accents and awkward questions.

What should I ask about data protection?

Ask where recordings and transcripts are stored, who can access them, how long they are kept, whether they are used to train models, and for the provider's data protection terms to read before signing. Legal requirements vary by country and sector, so check with a qualified adviser.

How do I avoid surprise charges?

Ask what the allowance measures, whether that is minutes, calls or conversations, and what happens when you go past it. Ask about setup fees and charges for changes, and get the answers in the contract or an email.

What are the biggest red flags?

Promises of perfect accuracy, vague answers about data storage, no unedited calls to hear, an unclear handover process, hidden fees, a long lock-in with no data export, and nobody named to support you after launch.