AI telephone answering service: costs and setup guide
An AI telephone answering service answers business calls, handles common questions and records each enquiry when your team is unavailable. The useful choice is not simply the service with the most features, but one that follows your real call flow, captures complete lead details and hands over urgent matters clearly.
What an AI telephone answering service should do
A practical service should answer using information supplied by your business rather than improvise. It should identify the caller's reason for calling, collect contact details, provide approved information and either book an appointment or pass the enquiry to your team.
For example, a plumber may need the caller's postcode, property type, fault, urgency and preferred visit time. A dental practice may need the patient's name, contact number, appointment preference and whether the enquiry concerns an existing booking. These fields should be agreed before the system goes live.
There are three common outcomes:
- Answer: the caller receives a straightforward response, such as opening hours, service areas or preparation instructions.
- Capture: the service records a lead for a follow-up call, quotation or consultation.
- Book or escalate: it schedules an available slot or sends an urgent matter to a person.
AI answering service versus the alternatives
| Option | Best for | Main limitation |
|---|---|---|
| Voicemail | Very low call volumes and non-urgent enquiries | Many callers will not leave enough detail, and follow-up depends entirely on your team |
| AI telephone answering | Regular enquiries, missed calls and out-of-hours cover | Needs accurate business information and a defined escalation process |
| Human receptionist | Complex conversations, sensitive cases and high-value sales calls | Usually involves staffing, shift or per-call costs and availability constraints |
| Call centre | Large volumes and broad support processes | Can be less personal and may require more configuration |
AI is most useful when callers ask repeatable questions. It is less suitable as the only option for emergencies, complaints requiring judgement or conversations involving confidential clinical or financial decisions. In those cases, the service should gather a message and route it to an appropriate person.
How much does an AI telephone answering service cost?
Compare plans by included conversations, phone minutes, languages, booking features, integrations and the cost of exceeding the allowance. A low monthly price can become expensive if every transfer, appointment or additional minute is charged separately.
As one straightforward example, Lyla plans start at £30 per month for about 300 conversations or £70 per month for about 1,000. Phone answering is an optional £40 add-on, with booking at £10, a custom voice at £20 and image understanding at £10. Setup and testing are free, with payment starting on go-live.
Before comparing quotes, calculate your likely volume. If you receive 20 calls on each of 22 working days, that is about 440 conversations a month. Also allow for evenings, weekends and repeat callers. Ask whether a conversation means a whole call, a message or a fixed number of minutes, because these definitions affect the real cost.
Questions to ask before choosing a provider
- What information does it use? Check that you can supply and update your own services, prices, locations, opening hours and policies.
- How are uncertain answers handled? The system should say when it cannot answer, capture the details and offer a clear next step rather than guess.
- Can it answer a dedicated number? Check whether you can keep your existing number, forward calls or use a dedicated number. Lyla provides dedicated numbers in 22 countries.
- What happens to every enquiry? Look for a transcript, caller details, time, reason for contact and an obvious notification for your team.
- Can it book safely? Confirm which calendar or booking process is supported, what availability rules apply and how cancellations are handled.
- How does it connect to existing tools? CRM, email and spreadsheet connections can prevent staff from copying information manually. Lyla connects to CRMs through Zapier and Make.
- What languages are supported? If you serve a multilingual area, check both understanding and the quality of spoken responses. Lyla supports more than 25 languages.
A practical setup process
1. List your top call reasons
Review a week or month of calls and group them into no more than 10 to 15 common reasons at first. Include new enquiries, pricing questions, appointment requests, cancellations, directions and urgent cases.
2. Write approved answers
Use short, definite answers. State where you work, what you do not do, your opening hours, response times and any qualification or access requirements. Mark information that must never be invented, such as availability, final prices or medical advice.
3. Define lead fields and escalation
Choose the minimum useful fields for each call. A trades business may need postcode and urgency, while a salon may need treatment, stylist and preferred time. Define words or situations that require a human response, then specify who receives the message and how quickly.
4. Connect the destination
Send captured enquiries to email, a CRM or another workflow. If you use Zapier or Make, test that the caller's name, number, message and source arrive in the correct fields. Do not assume a connection is working because the setup screen reports success.
5. Test real scenarios
Make test calls from different numbers and try accents, interruptions, vague questions, spelling out email addresses and requests outside your service area. Test during opening hours and outside them. Check the transcript, notification, booking result and escalation message for each scenario.
What to measure after launch
Review performance after the first seven and 30 days. Useful measures include answered calls, completed lead records, missed or abandoned calls, booking completion, transfers, repeat questions and the time taken to follow up. Also sample conversations for incorrect answers and missing details.
Set a simple target for each call type. For example, every quote request should contain a name, phone number, service required and postcode. Every appointment request should show the requested service and preferred time. If records regularly miss one field, change the question rather than asking callers to repeat themselves later.
When a human handover is essential
Use human cover for emergencies, safeguarding concerns, complaints, complex quotations and any matter where a wrong answer could cause harm. The AI service can still perform the first step by identifying the issue, collecting a safe callback number and explaining when the caller should expect a response.
A good AI telephone answering service is therefore not a replacement for every conversation. It is a reliable first point of contact that turns unanswered calls into structured, actionable enquiries while your team focuses on the work itself.
Frequently asked questions
Can an AI telephone answering service use my existing business information?
Yes. Choose a service that lets you provide your own services, locations, opening hours, policies and frequently asked questions. Keep this information current and test changes before relying on them for live calls.
Can it take messages outside normal opening hours?
Yes, if out-of-hours answering is included in the plan. Set a clear message for response times and define which situations should be treated as urgent rather than left for the next working day.
Will an AI answering service book appointments?
Some do. Check which calendar or booking system is supported, whether staff availability and buffers are respected, and how cancellations or changes are recorded.
Is AI telephone answering suitable for a small business?
It can suit plumbers, dentists, salons, clinics and other local services with repeatable enquiries and missed calls. It is less suitable as the sole response for sensitive or highly complex conversations that need professional judgement.
What happens if the caller asks something the system does not know?
The safest process is to say that the information is not available, capture the caller's details and create a follow-up task. Avoid services that encourage confident answers without showing how uncertainty is handled.