An AI receptionist for auto repair shops should finish one kind of call on its own, start a second kind and hand it over, and stay out of a third kind entirely. Owners tend to treat that as a software question. It is a sorting question. Before you compare products, you need a rule for which calls the machine completes, which it only opens, and which it never touches.
The pressure behind this is real. According to S&P Global Mobility, the average light vehicle on US roads reached a record 12.6 years old in 2024, and the figure rose again in 2025. Older cars produce repair orders that stay open for days: a part on backorder, a second failure found during teardown, an approval nobody has returned. Every open RO is a customer who will call to ask where things stand, and the person who picks up is the same service advisor who should be writing estimates and presenting inspection findings.
What follows is the decision tree I use to sort a shop's phone traffic. Run each call type through it once, write down the result, and you have your configuration.
The four-question sort for every call your shop receives
Ask these in order. Stop at the first one that sends the call to a person.
- Does the answer already exist as a field in your shop management system? RO status, promise time, parts ETA, open appointment slots, shop hours. If yes, continue. If the answer lives in someone's head or on the whiteboard by the parts counter, the call goes to a human until you fix that.
- Would a wrong answer cost money or create an authorization problem? Prices on vehicle-specific repairs, approval of added work, and warranty coverage all fall here. If yes, the AI collects details and hands off. It does not answer.
- Does answering require seeing, hearing, or driving the car? If yes, it is a diagnosis question. The AI books the car in and captures symptoms. It never names a cause.
- Is the caller describing a safety condition, or are they upset? If yes, transfer immediately, regardless of how the first three questions came out.
A call that clears all four belongs to the AI from greeting to goodbye. In most shops that describes exactly three things: status checks, appointment booking, and policy questions such as hours, shuttle availability, and night-drop instructions.
Wondering which AI tool pays for itself first? Request a free AI audit and find out.
Can an AI receptionist answer 'is my car ready' calls?
Yes, when it can read the repair order. A status check is a database lookup. The AI matches the caller to an open RO, reads the current status and promise time from the shop management system, and repeats what your advisor would have read off the same screen. It works only as well as your team keeps that status current.
First branch: is this the right caller?
- If the caller ID matches the phone number on an open RO, confirm the last name or the vehicle and proceed.
- If the number does not match (a spouse calling from work, a fleet manager calling about a driver's truck), ask for the name on the repair order and the vehicle. Two matching details, proceed. One or none, take a message for the advisor.
- If there are two open ROs under one customer, ask which vehicle before saying anything about either.
Second branch: what does the RO say?
- If the status is complete and the invoice is closed, the AI says the vehicle is ready, gives pickup hours, and explains after-hours key pickup if you offer it.
- If the status is in progress with a promise time, the AI gives the promise time exactly as entered. It does not estimate, round, or reassure.
- If the status is waiting on parts with an ETA, the AI says the part is on order, gives the expected arrival, and states that the promise time will be updated when it lands.
- If the status is waiting on customer approval, the AI tells the caller that the advisor sent an estimate or inspection report that needs a decision, offers to resend the link by text, and offers a transfer to the advisor. This branch is where the money is. A car sitting on a lift waiting for approval ties up a bay and a technician.
- If the status has not changed in longer than your threshold, or no RO matches, the AI does not guess. It says an advisor will call back with an update and sends the advisor a flagged message.
That last branch is the one that keeps the system honest. An AI that tells a customer the car will be ready at four, when the tech found a seized caliper bracket at two and nobody updated the ticket, has done more damage than a call that rang out. The handoff logic matters as much as the answers, and I have written separately about setting escalation rules so the AI hands off at the right moment.
Should an AI receptionist give repair estimates over the phone?
Only for services your shop already prices from a fixed menu. An AI receptionist can state the published price for an oil service, an alignment, or your diagnostic testing charge. Anything built from labor-guide hours and vehicle-specific parts goes to an advisor, with the vehicle details already collected.
Sort estimate calls into three branches.
- If the service is on your menu, the AI quotes it and books the slot, after asking year, make, model, and engine. Menu prices still carry qualifiers: a European car that takes eight quarts of an approved full synthetic is not your standard oil service, and the AI needs that rule written down.
- If the repair is vehicle-dependent, the AI collects year, make, model, engine, mileage, and the VIN if the caller has it, then promises an advisor callback with a written estimate. Front brakes, timing components, alternators, and water pumps all live here. Rear pads on a car with an electronic parking brake require a scan tool to put the calipers in service mode, which changes the labor against the same job on a cable-brake car. Whether an engine has a timing belt or a chain depends on which engine option is under the hood. A phone number given without that information is a number you will either eat or argue about.
- If the caller wants a price to fix a symptom, that is not an estimate request. It is a diagnosis request. Go to the next section.
There is also a legal reason to hold this line. In California, Business and Professions Code section 9884.9 requires a written estimate and customer authorization before work begins, and a separate documented authorization before any work beyond the original estimate. Florida's Motor Vehicle Repair Act requires a written estimate once a repair exceeds one hundred dollars. An estimate is a regulated document in those states, and many others have similar rules. Configure the AI so it never accepts approval for added work by voice. It resends the authorization link or transfers to the advisor, who records the approval the way your state requires.
One more item for the menu branch: keep the knowledge base current. Texas ended annual safety inspections for non-commercial vehicles on January 1, 2025, while emissions testing continues in designated counties, including Travis and Williamson. Callers still ask whether you do state inspections. An AI running on last year's answers will book appointments for a service that no longer exists.
Why do diagnosis questions at a repair shop still route to a human?
A diagnosis is a judgment made with the car in the bay, and no phone description contains enough information to make it. A symptom maps to many causes. An AI receptionist that names one has made a promise your technician must either keep or contradict. The correct job for the AI is intake: capture the symptoms well and get the car scheduled for testing.
Take the caller who had the code read at a parts store and wants a price for a catalytic converter because the printout said P0420. That code means catalyst efficiency below threshold. The converter may have failed. So may the downstream oxygen sensor. There may be an exhaust leak ahead of the sensor, or the engine may be running rich and overheating a converter that a replacement will not save. A shop that quotes a converter over the phone and then installs one on a car with a leaking flex pipe has created a comeback and probably a one-star review. Your advisor knows to say the code is where testing starts. The AI should say the same thing in simpler words and then book the appointment.
Here is what the AI should collect on a diagnosis call, so the tech starts with a real complaint line on the RO instead of 'customer states noise':
- When it happens: cold start, highway speed, braking, turning, over bumps, with the A/C on
- Which warning lights are on, and whether the check engine light is steady or flashing
- How long it has been happening and whether it is getting worse
- Any recent work, at your shop or elsewhere
- Whether the vehicle is drivable or needs a tow, and whether the customer plans to wait or drop off
The immediate-transfer list
Some phrases skip the tree entirely. If the caller says any of the following, the AI transfers to a person right away, and if no one is available it gives your towing instructions:
- Brake pedal goes to the floor or the car will not stop normally
- Check engine light is flashing, which signals an active misfire that can destroy the converter
- Temperature gauge in the red or steam from under the hood
- Smell of fuel
- Steering that suddenly became heavy or loose
- Any complaint about work your shop just completed
That last one is not a safety item, but it is a relationship item. A customer calling about a possible comeback should reach a person who can own it.
Is your shop ready for an AI receptionist? Three gates to pass first
The tree above assumes things that are not true in every shop. Check these before you turn anything on.
Gate one: status discipline. Pull five open ROs at two in the afternoon and compare the status in the software to what is physically happening in the bays. If they match, proceed. If the whiteboard is the truth and the software is a day behind, fix that first. Otherwise the AI will read stale data to customers with complete confidence.
Gate two: read access. Ask any vendor to demonstrate a status lookup against your own system, whether that is Tekmetric, Shopmonkey, Shop-Ware, Mitchell 1 Manager SE, or something else. If the tool can read RO status and promise time, the full tree applies. If it cannot, the AI is still useful for booking, intake, and after-hours calls, but status checks stay with your advisors. In that case, move the effort to proactive text updates at each status change. The best status call is the one the customer never needs to make.
Gate three: consent and disclosure. Inbound calls are simple. Outbound is not. In February 2024 the FCC ruled that AI-generated voices count as artificial voices under the Telephone Consumer Protection Act, so an AI voice that calls customers to say their car is ready needs prior consent. Add consent language for calls and texts to your drop-off authorization form. If you record calls, have the greeting say so; California and several other states require every party to consent to recording.
If you pass all three, the remaining work is writing down what the AI says in each branch. The structure in this guide to AI receptionist scripts carries over well. Just substitute your branches for the generic ones.
What should an auto repair shop measure in the first 30 days?
Count your calls by type for one week before launch, even if it is tally marks on a legal pad at the counter. Then track these after:
- Status calls reaching an advisor. This should fall sharply. If it does not, callers are asking for a person, which usually means the answers are too vague or the data is stale.
- Wrong-status incidents. The target is zero. Each one traces back to gate one.
- Time from estimate sent to estimate approved. If the waiting-on-approval branch works, this shortens, and bays turn faster.
- After-hours appointments booked. These are jobs that previously went to voicemail or to the next shop in the search results.
- Average repair order. It should hold steady or rise as advisors get time back to present inspection findings properly. If it drops, the AI is answering questions an advisor should be handling.
Listen to twenty recorded calls a week for the first month. You are listening for one failure: the AI answering a question that the four-question sort should have sent to a human. Tighten the rule each time you hear it. If you are still deciding between software and people on the phones, the comparison of an AI receptionist against a live answering service covers that trade-off, and our AI receptionist overview describes how Turnkey AI sets these up for service businesses.
The whole tree on one page
- Status check, RO matched, status current: AI answers from the repair order.
- Status check, RO stale or unmatched: AI takes a message and flags the advisor.
- Waiting on approval: AI resends the link and offers the advisor. It never accepts approval by voice.
- Menu-priced service: AI quotes and books.
- Vehicle-dependent repair price: AI collects vehicle details and the advisor returns a written estimate.
- Symptom or code: AI collects the complaint and books testing. It names no cause and gives no price.
- Safety phrase, comeback, or upset caller: immediate transfer.
None of this takes judgment away from the people who have it. Your technicians still diagnose. Your advisors still build estimates, explain findings, and earn the approval. The AI takes the lookups, so the phone stops interrupting the work only they can do.
Questions shop owners ask before turning it on
Can an AI receptionist look up a repair order in Tekmetric, Shopmonkey, or Mitchell 1?
It depends on the specific product and the access your shop management system allows. Ask the vendor to show a live status lookup against your own account before you commit. Without read access to RO status and promise time, the AI can still book appointments and take intake, but it cannot answer status checks accurately.
Can the AI approve additional work when a customer says yes on the phone?
It should not. Many states require documented customer authorization before work beyond the original estimate, and California's Automotive Repair Act is explicit on this. Have the AI resend the digital authorization link or transfer to the service advisor, who records the approval properly.
Will customers hang up when they realize they are talking to AI?
Some will ask for a person, and the AI should offer that transfer at any point without resistance. A caller who only wants to know if the car is ready generally cares more about a fast, correct answer than about who delivers it. Watch your transfer requests in the first month and adjust.
What does the AI do with after-hours calls and tow-ins?
It books appointments into open slots, explains night-drop and key-drop procedures, gives your preferred towing instructions, and captures the complaint so the RO is half-written when the advisor arrives. Safety-related calls get your emergency guidance and a priority flag for the morning.
Does an AI receptionist replace a service advisor?
No. It removes the lookup calls from the advisor's day. Estimates, diagnosis conversations, inspection presentations, and comeback handling all require a trained person, and those are the conversations that determine what a shop earns.
Frequently asked questions
Can an AI receptionist look up a repair order in Tekmetric, Shopmonkey, or Mitchell 1?
It depends on the specific product and the access your shop management system allows. Ask the vendor to show a live status lookup against your own account before you commit. Without read access to RO status and promise time, the AI can still book appointments and take intake, but it cannot answer status checks accurately.
Can the AI approve additional work when a customer says yes on the phone?
It should not. Many states require documented customer authorization before work beyond the original estimate, and California's Automotive Repair Act is explicit on this. Have the AI resend the digital authorization link or transfer to the service advisor, who records the approval properly.
Will customers hang up when they realize they are talking to AI?
Some will ask for a person, and the AI should offer that transfer at any point without resistance. A caller who only wants to know if the car is ready generally cares more about a fast, correct answer than about who delivers it. Watch your transfer requests in the first month and adjust.
What does the AI do with after-hours calls and tow-ins?
It books appointments into open slots, explains night-drop and key-drop procedures, gives your preferred towing instructions, and captures the complaint so the RO is half-written when the advisor arrives. Safety-related calls get your emergency guidance and a priority flag for the morning.
Does an AI receptionist replace a service advisor?
No. It removes the lookup calls from the advisor's day. Estimates, diagnosis conversations, inspection presentations, and comeback handling all require a trained person, and those are the conversations that determine what a shop earns.