In many garages the phone rings in the workshop, or at a desk where the same person writes the estimates, orders the parts and hands cars back to their owners. An AI voice assistant for a garage answers the calls that would otherwise be lost: it asks what car you have, what is wrong with it and when you can come in, then puts the booking in the workshop diary.
It introduces itself from the outset as an automated assistant and has a narrow role: it gathers information and puts it in front of the people who make the decisions. It does not judge what is wrong with the car, does not say what the repair will cost and does not promise when it will be ready.
Who answers when the receptionist is out on the forecourt, handing over a car
A garage’s calls come in waves: in the morning, when cars are dropped off, and in the afternoon, when owners want to know whether they can collect them. That is exactly when reception is busy with handovers and inspections. Anyone who cannot get through rings somewhere else, especially for a service or a tyre change.
There is another loss, a less visible one: the booking taken in a hurry. “I’ll come in on Thursday with a Logan, it’s making a noise” tells the mechanic neither which engine the car has nor which parts might be worth ordering in advance.
AI voice assistant for garages: the check-in sheet taken over the phone
The assistant goes through the questions an attentive service adviser would ask. The description of the fault stays in the driver’s own words, with no interpretation: “it squeals under braking, at the front, worse in the morning” is more useful than a category picked by a program.
It draws no conclusions from the answers, it just enters them on the sheet. If the caller asks whether they can keep driving the car, the assistant offers no opinion and puts them through to the adviser.
- The car: make, model, year of manufacture, engine and registration number.
- The symptoms: what can be heard or felt, since when, in what conditions, which warning lights are lit on the dashboard.
- The type of job: service, brakes, tyres, diagnostics, bodywork, so that the booking goes to the right bay.
- The booking: it offers the free slots in the workshop diary and sends the confirmation by text message.
- Callback: it takes the name, number and subject and adds the request to the adviser’s list.
“Is the car ready?”: job status, read from the garage’s software
The second big category of calls comes from people whose car is already in the workshop. The assistant identifies the caller by phone number and registration number, finds the open job card and reads out its status: waiting, in progress, awaiting parts, ready for collection.
It reads, it does not estimate. The quality of the answer depends on how up to date the workshop keeps the statuses in the software: if the mechanic changes the status only in the evening, the assistant will spend the whole day saying something out of date. This is a matter of internal discipline to settle before going live.
No diagnosis, no price, no promised date
The driver will ask anyway: “What do you think is wrong with it?”, “How much will it cost me?”, “Will it be ready by Friday?”. To all three the assistant gives the same honest answer: it has no way of knowing, and the answer comes from the adviser once the car has been seen. A diagnosis made over the phone by a program would be guesswork, and a price spoken by it would be taken as a quote.
Like any speech recognition system, it can mishear, especially a registration number dictated on the move. That is why it repeats the details and asks for confirmation, and when the conversation gets stuck it transfers the call or logs a callback request.
Connecting to the garage software and the trial period
The garage keeps its phone number; the existing phone system sends the assistant the calls you choose. Speech becomes text through speech recognition, a language model (the software that understands the request) follows the agreed scripts, and speech synthesis says the answer aloud. Bookings and statuses travel between the assistant and the garage software through an API, that is, an interface through which two applications send each other data.
If your software has no such interface, the assistant works with a separate calendar and sends the sheets by email, and job status remains a question for a person.
- We listen to how people talk at reception: what drivers ask and what is usually missing from the sheet.
- We set the capacity: how many cars come in per day for each type of job.
- We write the refusals: what the assistant says when it is asked for a price, a date or an opinion.
- We test with calls from the workshop, with compressor noise and with the phone on speaker.
- We start with missed calls and out-of-hours calls, then extend.
- We tell the caller if the call is being recorded and decide how long the data is kept, in line with GDPR.
Frequently asked questions
What happens if the assistant does not understand what the caller says?
It asks once more, and if things are still not clear it transfers the call to reception or logs a callback request. It does not fill in the sheet with assumptions.
Can it say how much a repair costs?
No. The cost depends on the inspection, the parts and the labour, which the adviser works out. The assistant notes the question and tells the driver that the estimate will come from a person.
How long does it take before it can answer the phone?
It depends mostly on the garage software: if it already has an interface for bookings and job cards, the connection is quicker to set up. It also matters how many types of call you want it to handle from the start.
This is a typical project description: it shows how we usually approach this kind of work and does not present a project carried out for a particular client. Every real project starts from your company’s situation, and the stages, timescales and price are agreed after the initial discussion.