AI for Mechanics Reads the Fault Code. The Money Dies on the Job Card

At 6:40 on a Friday night, Tony is at the counter of his four-bay workshop in Ipswich, flicking through a stack of finished job cards and wondering why a flat-out month felt thin at the bank.

His new software had a great month. The phone assistant catches the calls that used to ring out while he was under a Hilux. The booking system fills Monday from the weekend's enquiries. The scan tool's diagnostic assistant turns a two-hour electrical mystery into a forty-minute job.

None of that helped with the Camry that left in May with the front pads at three millimetres and a polite "we'll do it next time". Or the ute with the inside edge of both front tyres going, noted on the card, never mentioned again. Or the forty minutes his apprentice spent chasing a rattle that never made it onto the invoice.

Nobody decided to give that money away. The workshop simply found more work than anyone had time to chase. That is the gap in how AI for mechanics is being sold: diagnostic assistants, booking bots, phone answering, inspection apps with a camera and a confidence score. All of it aimed at the car in the bay today. I build these systems for Australian businesses, so let me say the uncomfortable half out loud. Most workshops are not short of cars. They are short of the second visit that the first visit already found.

What AI for mechanics gets right

Give the front-of-shop tools their due, because the case for them is real.

A workshop loses jobs at the phone, not the hoist. An assistant that picks up while every human is elbow-deep in an engine bay is doing work nobody was going to do anyway. The diagnostic helpers are genuinely good too: pattern-matching a fault code against a few million prior repairs is exactly what a model is for, and it makes a young tech faster without making anyone reckless. I wrote about the sole-trader version of this in AI for tradies, and it holds for a shop with a driveway.

But notice what all of it touches: the car in the door today. The software industry builds there because a booked job is the one thing that is simple to count.

The half of AI for mechanics nobody demos

Now walk the same shop from the job-card drawer instead of the driveway.

Every inspection your techs do produces findings. Some get approved at the counter. The rest get declined, softly: not today, after the school holidays, when the rego is due. The finding lives on a paper card or a free-text note in your workshop software, and that is where it dies. Nobody rings the Camry in September, because September's phone is ringing about September's cars.

The follow-up call is the most profitable call a workshop can make, and the first thing dropped at 5pm on a Friday.

A job card is a list of reasons the car will be back. Nobody in the building is paid to collect them.

The job-card maths nobody has time to do

Run your own shop as arithmetic.

Say forty cars cross the hoists in a week, your techs flag something on half of them, and half of those findings get approved on the day. That leaves ten declined items a week. Call the average one $380: pads and rotors, two tyres, a coolant flush, a leaking rocker cover. That is $3,800 a week driving out of the car park with a due date attached.

Nobody converts all of it, and nobody should try. But if a follow-up that quotes the tech's own note brings back one in five, that is around $760 a week, close to $38,000 a year, from work you already found, quoted with hands you already paid.

The fix is not a better sales pitch. It is remembering, four months later, what your tech wrote down, which is exactly what a busy shop cannot do and a system cannot forget.

Start with the declined work, not the diagnostic bot

Pick one workflow, not the whole workshop.

A system reads the job cards and inspection notes already sitting in the software you run. It holds every declined item with the date, the odometer, and the exact words your tech wrote. It knows pads at three millimetres in May are due about now, and it drafts the message in your shop's voice: what was found, the photo from the inspection, a price, a link to book. On Monday morning someone at the counter reads the batch and taps approve or skip. Nothing goes out on its own.

Your techs keep finding the work. The counter keeps the judgement. The chasing stops depending on anyone's memory.

That gives you two numbers a month: items chased, and cars rebooked. The distance between what your techs find and what your invoices show stops being a Friday-night feeling and becomes a report.

Then point the same loop at your parts. Supplier invoices say what came in the door; customer invoices say what went out on cars. The gap is parts you paid for and never billed. A system reconciles the two line by line and hands you the exceptions.

Automate the reminder, never the spanner

Here is the line that keeps this safe.

No system decides whether a car is roadworthy, whether those pads can wait until the holidays, or what to tell a customer asking if the car is safe to drive. Those calls sit under your name and, for a roadworthy, under a licensed examiner's signature. A reminder must quote what your tech actually wrote, never inflate it: a text that says unsafe when the card says worn is how a workshop loses twenty years of trust in one afternoon.

Aim the software at the job cards and the phone, never at the hoist.

What AI for mechanics costs

Work down this list in order and stop the moment something works.

• The tools already inside your workshop software. Service reminders, inspection templates, SMS follow-ups. Usually half configured. Turn them on before you spend anything.
• A subscription tool for digital inspections or follow-up, priced per shop each month. Cheap enough to trial for a quarter and judge on rebooked cars.
• A custom build that reads your job cards, drafts the follow-ups, reconciles the parts and writes back into the software you already run. A few thousand up to the mid teens of thousands, depending on how many systems it must talk to.

Hold that against ten declined items a week. What catches shops out is the running cost rather than the build, and I broke those layers apart in what AI actually costs a small business.

The Australian layer: warranties, roadworthies and rego data

Two things separate a system built for an Australian workshop from an overseas template.

The first is the rules around what you say. A dealer cannot void a new-car warranty because the logbook servicing happened at your shop, and your follow-ups can say so plainly, but a message that overstates a safety finding walks into misleading-conduct territory under consumer law. The safe pattern is boring: quote the card, show the photo, state the price.

The second is data. Your system holds names, phone numbers, registration plates, addresses and service history. Ask where it is stored, whether it trains someone else's model, and whether you can get it out when you change platforms.

Common questions about AI for mechanics

What should a workshop automate first?

Declined-work follow-up. It runs on job cards you already write, and one quarter measures it with two numbers: items chased and cars rebooked.

Will AI replace mechanics?

No. Hands on the car and judgement about the car are the product. The remembering and the ringing around it are not.

Is an AI diagnostic assistant worth it?

Often, if diagnosis time is genuinely your bottleneck. Check that first. If your problem is found work leaving unchased, diagnosing faster just finds more work to forget.

Go back to Tony. He did not have a quiet month. He had a month where his techs found thousands of dollars of honest, needed work, and the only copy of it went into a drawer under the counter.

If you want a straight read on what your job cards are worth, that is what a first call is for. Book a strategy call and bring one month of job cards and the invoices that followed them. We will count the declined work before we talk about building anything.