← Writing
What AI can actually do for a small auto repair shop
Every software vendor now says "AI". For a 2–5 bay shop, the question is simpler: what will save my service writer an hour a day, and what will get me in trouble with a customer? I build shop software, so here's my honest list.
Jobs AI does well today
- Turning notes into the three C's. A technician dictates "front pads are down to two millimeters, rotors scored, customer hears grinding" and AI writes a clean concern, cause and correction. The tech checks it in ten seconds instead of typing for two minutes.
- Explaining a repair to the customer in plain words. From what the technician recorded, AI can draft two or three sentences a customer actually understands — "your brake pads are worn thin, which is why you hear grinding" — without jargon. This only works if the shop approves every explanation before the customer sees it.
- Reading paper. A photo of a parts invoice becomes inventory and a purchase. A photo of the odometer fills the mileage. A picture of a part number fills the line.
- Suggesting labor and parts. When a writer starts "front brake pads and rotors" on a 2012 Camry, the system can suggest hours and parts — ideally from the shop's own history first, so the suggestion matches how that shop actually works.
- Following up. Remind the customer about the work they declined last month, at the right time, in a message that sounds like the shop.
Jobs AI should not do
- Set prices on its own. AI can suggest; a person decides. A wrong price sent to a customer is a trust problem, not a software problem.
- Invent urgency. "This is dangerous, fix it today" must only come from the technician. Good systems block the AI from adding urgency the tech didn't record.
- Write repair procedures or torque specs. Those come from manufacturer data, not from a language model.
- Send anything to a customer without a person seeing it first.
The rule I build by: AI drafts, a person decides. Every action comes back as something the shop confirms or edits.
Where a small shop should start
- Text-to-approve estimates. Not AI, but the biggest time saver: the customer approves job by job from a link, signs with a finger, and the shop stops chasing phone calls.
- Dictated notes into the three C's. Low risk, saves typing every single job.
- Customer explanations with approval. Fewer "what does that mean?" calls, more approved work.
- Receipts to inventory. Once parts invoices are photographed instead of typed, inventory stays honest.
Questions to ask any vendor
- Can a person review everything the AI writes before the customer sees it?
- Where do the labor and parts suggestions come from — my history, or a generic guess?
- If the AI fails, do I still pay for it?
- Does it work on a phone in the bay, not only at the front desk?
I built these ideas into ORAVA, made in New Jersey for independent shops. If you'd like to see it, or talk about AI in your shop, write to me: israel@aquilapriscilla.com