Garages use AI to turn a technician's shorthand into a plain-English text at four moments: check-in, when a fault needs the customer's approval, when parts or labour time slip, and when the car is ready. Customers appreciate updates that arrive before they have to ring, explain each finding in one line, and include a photo.
The AI's job is translation and tone, not judgement. Whether a fault is urgent, what it costs and whether the car is safe to drive all come from the technician and the estimate in your garage management system. Most "is it ready yet?" calls come from silence rather than bad news, so the biggest win is usually sending something at all, and sending it before the customer wonders.
The four repair updates drivers actually want
Ask customers what annoyed them about a garage visit and the answers are rarely about price. They're about not knowing. These four messages cover almost every question a customer would otherwise phone about:
| Moment | When to send | What it must say | Length |
|---|---|---|---|
| Checked in | Within an hour of drop-off | Car is booked in, who's looking after it, when you'll next update | 2 sentences |
| Approval needed | As soon as the inspection is written up | What was found, why it matters, the price, what happens if they say no | Up to 6 short lines, plus photos |
| Delay | The moment you know, not at collection time | What changed, the new time, what you're doing about it | 3 sentences |
| Ready | When the car is parked and the invoice is checked | Work done, total, collection hours, any advisories for later | 4-5 lines |
Notice the timing column. A "ready" message sent while the car is still on the ramp is worse than no message, because the customer arrives and waits. The trigger has to be a status your team sets deliberately, not the technician closing a labour line.
From technician shorthand to a message a driver understands
Technicians write for other technicians, and they should. The problem is when that note is copied into a customer text unchanged. Here's an illustrative before and after for the approval message.
Technician's note: FL lower arm balljoint excessive play. Rear pads 3mm, discs lipped. Aux belt cracking. Coolant slightly low, no visible leak, topped up.
Illustrative AI rewrite for the customer:
Hi [first name], we've inspected your car. Three things to decide on:
RED, needs doing now: the front-left suspension joint is worn and loose. This affects steering and should be replaced before the car goes back on the road. $185 fitted.
AMBER, soon: rear brake pads are nearly worn out and the discs have a ridge. Fine for a few weeks, but we'd do both together. $240 fitted.
AMBER, soon: the engine drive belt is cracking. $95 fitted.
GREEN, done: coolant was a little low, so we topped it up. No leak found.
Photos attached. Reply YES to all, or tell us which items. Questions? Call [adviser's name] on the workshop line.
Two things make that work. It groups findings by urgency, which is the traffic-light layout most digital vehicle inspection (DVI) tools use: a digital checklist where the technician records each item with photos and a red, amber or green rating. And it keeps the technician's judgement intact. The AI didn't decide the ball joint was red; the technician did. What the AI added was "affects steering", which is the sentence that gets a customer to say yes.
What you'd check before sending: that every price matches the estimate line in your system, that "should be replaced before the car goes back on the road" reflects what the technician actually said about safety, and that nothing was quietly dropped.
It helps to give the assistant a short glossary of your technicians' usual shorthand, so the translation is consistent from one adviser to the next. A starting set, which your own technicians should check and extend:
| Technician writes | Customer reads |
|---|---|
| CV boot split | The rubber cover over a driveshaft joint is torn, so dirt and water can get in and wear the joint |
| Discs lipped | The brake discs have a ridge worn into the edge, so new pads won't sit flat against them |
| Track rod end play | A joint in the steering is worn and loose |
| Aux belt cracking | The drive belt that runs the alternator and other engine parts is cracking |
| DPF light on | The warning light for the exhaust filter that traps soot is on; it needs checking |
| Wipers smearing | The wiper blades are worn and leave streaks |
Paste the glossary under the prompt's rules with the line "Use these translations where they apply." When a technician uses a new abbreviation and the draft guesses at it, add the right translation that day.
How the AI step fits into a garage's existing software
There are two ways this usually works in a small garage.
Inside your garage management system. Several platforms now build AI into the inspection and messaging flow. Tekmetric announced an AI "Smart DVI" in April 2026 that turns a technician's spoken video walk-round into a written inspection with organised photos and estimate items; it was described as coming soon at launch, so check whether it's live on your account. Others, such as Shopmonkey, pair two-way texting, digital estimate approvals and inspection photos with message templates, and use AI in places like suggested replies to reviews. Check what your own platform includes. If your system has an AI drafting feature, start there, because it already knows the customer, the vehicle and the prices.
Alongside it. If your system doesn't, a service adviser can paste the technician's notes and the estimate lines (no customer phone numbers or addresses needed) into a business chat assistant, get the rewrite, check it, and send it through the system's normal texting. It's a copy-and-paste step, but it takes about a minute a job, and it works with any software.
Either way, the flow is the same: technician records findings, AI drafts, adviser checks prices and wording, message goes out. Skipping the adviser check is the step that causes trouble.
A prompt service advisers can reuse on every job
If you're using a general assistant, save this as a reusable instruction (a Project in ChatGPT or Claude, or a Gem, becoming a skill from November 2026 in Gemini) so every adviser gets the same output.
You turn car technicians' inspection notes into customer texts
for a small independent garage.
Rules:
- Group items as RED (needs doing now), AMBER (soon), GREEN (fine / done).
- Use the rating the technician gave. Never change a rating.
- One plain-English line per item: what it is, why it matters.
- Use prices exactly as given in the estimate lines. Never
estimate, round or add a price that isn't listed.
- If the technician says a fault affects safety, say so plainly.
Do not soften it, and do not exaggerate items that aren't safety-related.
- No jargon: explain parts by what they do ("drive belt", not "aux belt").
- End with how to approve (reply YES or name items) and a named contact.
- Under 120 words.
Customer first name: [name]
Technician notes: [paste]
Estimate lines: [paste]The check-in message is short enough that a template usually beats AI, but if you do draft it, watch the promise it makes. An illustrative first draft: "Hi [first name], your car's with us and we'll call you by 12 with an update." Nobody had told the assistant when the inspection was scheduled; it filled the gap with a plausible time, and on a busy morning that inspection wouldn't start until 11. The fixed version takes the time from the booking: "Hi [first name], your car is booked in with [technician's first name]. We'll text you after the inspection, which we expect to finish by [time from diary]." A promise the garage misses by an hour does more harm than a message without one.
The line about not exaggerating matters as much as the one about not softening. Customers who feel pushed by an alarming message decline everything and don't come back. If you want to follow up on the items they declined, that's a separate, gentler sequence, covered in following up unapproved repair estimates with AI.
The delay message nobody enjoys sending
Delays are where garages lose goodwill, mostly because the message goes out at 4:45pm when the customer is already on their way. The AI won't fix the delay, but it removes the excuse that writing the message takes too long. An illustrative before and after:
Before (what often gets sent): "Parts delayed. Will update tomorrow."
After (adviser's facts, AI's wording): "Hi [first name], sorry, the replacement water pump for your car hasn't arrived with this afternoon's delivery. The supplier now says 9am tomorrow, and we've kept a technician free to fit it straight away, so we expect it ready by 1pm. If you need to get about in the meantime, let us know and we'll see what we can do."
The adviser gave the assistant three facts: the part, the new delivery time and the plan. The AI supplied the apology and the order. It must not invent the reason ("due to high demand") or the offer. In the example, "we'll see what we can do" is deliberately vague because the garage hadn't confirmed a courtesy car; if yours has one to offer, say so explicitly in the prompt.
A four-bay garage's Tuesday, before and after
To show the effect, here is an illustrative independent garage with four bays, three technicians and one service adviser, handling about 14 cars a day.
Before. The adviser sends texts only when approval is needed, typing them by hand between phone calls. Around 25 calls a day are customers asking for a status. Approvals often wait until the customer rings back, so a car found needing work at 10am might not get a yes until 3pm, and the bay sits idle or the job slips to tomorrow.
After. Check-in texts go automatically from the booking system when the car is marked "on site". Approval texts are drafted by AI from the technician's inspection within 15 minutes of it being written up, checked by the adviser in about a minute each, and sent with photos. A "ready" text goes when the adviser marks the invoice checked.
A realistic result would be status calls falling from about 25 to under ten a day, and most approvals coming back within an hour because the message is clear and has photos. At three or four minutes per status call, that frees over half an hour of the adviser's day, and faster approvals let the workshop finish more jobs the same day. The running cost is small: a chat assistant seat at around $20-$25 a month if your system doesn't include AI, plus whatever your platform charges for texting. The numbers are illustrative; count your own calls for a week before and after.
Where AI-written repair updates go wrong
A few misconceptions come up repeatedly when garages start using AI for updates:
- "The AI can tell customers what's wrong." It can only rephrase what the technician found. If you paste a vague note ("noise from front"), you'll get a confident-sounding message about a problem nobody has diagnosed. Garbage in, polished garbage out.
- "Friendlier is always better." Softening safety wording is the dangerous failure. "You may want to keep an eye on your brakes" for pads at the metal is a liability. Keep the technician's safety wording and let the AI explain it, not dilute it.
- "It knows our prices." A general assistant doesn't. It will happily write "around $200" if the estimate line is missing. That's why the prompt forbids estimating.
- "Automate the ready text on job completion." The technician finishing labour isn't the same as the car being washed, parked, invoiced and the keys on the board. Trigger it from a status the adviser sets.
- "More detail builds trust." Up to a point. A ten-paragraph message gets skimmed and the approval gets missed. Keep one line per item, and let the photos do the rest.
An illustrative failure worth describing: a garage switches on automated updates, and a customer gets an AMBER message about tyres "nearing the legal limit" when the technician had written 4mm, which is nowhere near it. The AI had filled a gap with a stock phrase. The fix was one rule in the prompt ("state measurements as given; don't add comparisons") and the adviser check that should have caught it.
Partial approvals cause a different mismatch. The customer with the three findings above replied: "yes to the steering joint, leave the brakes for now, and how long would the belt take?" The adviser answered the question and booked the joint and the belt, but didn't update the estimate in the system. At collection, the "ready" text was drafted from the original approval message and listed all three items with a total of $520, while the invoice correctly said $280. The customer rang, understandably worried about being charged for brakes they'd declined. The rule since then: the "ready" message is always drafted from the checked invoice, never from earlier messages, and it names what was left for later.
Done that way, the ready text reads as a clear record rather than a sales reminder. An illustrative before and after for the same car:
First draft: "Your car is ready. We replaced the front steering joint and drive belt. We did NOT replace the rear brakes, which still need attention. Total $280."
Sent: "Hi [first name], your car is ready to collect until 6pm. We replaced the front-left suspension joint and the drive belt; total $280, as agreed. You chose to leave the rear brakes for now; they were fine for a few weeks when we checked, and we'll send you a reminder in three weeks. Keys are at the desk."
The capitals and "still need attention" read as a telling-off for a decision the customer was entitled to make. The sent version records the same facts, keeps the technician's timescale, and says what happens next.
How to tell customers appreciate the updates
You'll know within a month whether it's working. Track these four things for two weeks before you change anything, then again after:
- Status calls per day. Tally on a sheet by the phone. This is the clearest signal.
- Time from inspection to approval. Your garage software records both timestamps.
- Approval rate on amber items. If it drops, your messages may be reading as pushy; if it rises without complaints, they're reading as honest.
- What reviews mention. "Kept me updated" and "sent photos" are the phrases you're hoping to see. Complaints about "being bombarded with texts" mean you've added too many messages.
If you're still deciding whether AI belongs in your workshop at all, the decision checklist for small auto repair shops is a good first pass. Once updates are running, the natural next step is bringing customers back for their next visit, which AI service reminders for garage customers covers. And if the adviser check feels like a bottleneck, setting up human review without slowing down shows how to make it a one-minute habit rather than a queue.
Further reads
- Writing Client Updates and Snag Lists With AI for Builders — The same plain-English update approach for building jobs.
- Turning On-Site Voice Notes Into Job Sheets and Quotes With AI — Turn a technician's voice note into a tidy job sheet.
- How to Use AI to Respond to Google Reviews — Reply to the reviews good updates tend to produce.
- Cutting Parcel-Tracking Calls With AI Delivery Updates — How couriers cut 'where is it?' calls with proactive updates.
- How to Add Human Approval Steps to AI Automations — Build the adviser check into any automated message.
- What Does an AI Implementation Consultant Actually Do? — The six stages of an AI implementation consultant's work, the evidence each stage should leave behind, and how the role differs from strategy or IT help.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Tekmetric announcements at Tektonic 2026 (Smart DVI, reported by Ratchet+Wrench); Shopmonkey customer communications and Reviews Manager pages (checked September 2026).