AI assistant for a car service
It answers the price question with a range for that particular car, books onto a bay that is genuinely free, and comes back when the next service falls due.
- Buildfrom $1,700
- Timeline2–4 weeks
- Supportfrom $570/mo
Who this suits: Shops where the advisor answers messages between two cars, the price question arrives all day, and customers come back only when something is already broken.
What is in the first version
- booking a bay
- price range
- service reminders
- advisor handover
The figures above are the package this sits on: AI agent or process automation. Anything past its edges the calculator adds before the work starts, not after.
How the path goes
- 01
The customer asks what it costs
From the car and the job it names a range from your price list, final only after an inspection.
- 02
The booking takes a real bay
The slot follows the bay load, the real length of the job and whether the part is in stock.
- 03
The car comes back on schedule
Mileage and date are recorded, and each owner is offered a specific time for their own car.
What decides whether this works or annoys people
A quoted price becomes a promise
The range comes with the condition that voids it, and anything above it is agreed before the work starts.
A diagnosis from a described knock
The assistant never names the part: it collects the symptoms, books a diagnostic and passes the description on.
A service reminder becomes spam quickly
The due date comes from mileage and date together, an opt out sits in every message, and an answer stops it.
What to measure once it is live
- bookings made without the advisor
- bay load across the week
- owners who come back on a reminder
- the gap between the range and the invoice
Why this comes out faster
The shape of this one is known: the states, the edge cases and the things that usually go wrong have been decided before. Nothing here is a template, and the saving is not in your half of the work. It goes into your process, your content and the systems this has to talk to, which is the part nobody can have solved in advance.
A likely stack for this
Picked against the task when we scope it, not decided in advance. This is the shelf it usually comes off.
- Assistant and serverPythonFastAPIPostgreSQLRedisTelegram Bot API
- Models and knowledgeOpenAIAnthropicpgvector
- Schedule and operationsn8nTemporalDockerGrafana
Price it yourself, right here
Five steps, and you can see the number without leaving a contact. The estimate accounts for the kind of work, what you already have and what it will need inside.
These bills do not come from us
- AI model calls at the provider’s rate
- Paid access to VIN and labour-time catalogues
- SMS reminders for owners outside messengers
- The plan on your shop software
Asked before the first call
- Where does it get the prices?
- From your own table of jobs and labour hours. Outside it nothing is named: on an unknown job it says it will check with the mechanic and passes the question on.
- What happens when a customer does not turn up?
- A reminder the day before and on the day, with a one-tap way to move it, and the freed bay becomes bookable again straight away.
- Our bookings live in our own software. Will it work with that?
- Yes: the calendar stays in your software, the assistant reads the load and writes the booking back. We agree in advance which one owns the schedule.