Automotive AI service schedulers,
put to the test
The same discipline as every Proofground guide: recorded calls, blind judges, published evidence — on the scenarios this trade lives on.
Report cards
Graded on the calls auto service departments actually get — recorded, judged against a written rubric, independently re-reviewed.
No grades published here yet. The test suite is built and waiting; we publish a grade only with the per-conversation evidence behind it.
Graded so far: DaVoice AI. Open a card for the letter, the calls behind it, and the date they were made — every grade is a dated snapshot, not a standing verdict, and the cards above are the current list.
What we test in a service drive
Service callers want two things the phone cannot honestly give them: a diagnosis and a price. A good scheduler books the diagnosis without pretending to be a technician, captures a VIN and an odometer without a transposed character, and recognises the caller who should not be driving the car at all.
Every vendor here runs the same 27-scenario suite, scored against the same rubric — a grade is only comparable because the calls are. What those scenarios cover:
- The everyday calls: the routine requests that make up most of a day; if these fail, nothing else matters. 3 tests — “Hours, location, and ‘do you work on my make?’” · “Returning customer moves an existing service appointment”
- Calls that change halfway through: the caller wants one thing, then another; both have to land. 3 tests — “Calling to cancel, then deciding to reschedule instead” · “A pricing question that turns into a booking mid-conversation”
- Getting the details exactly right: names, numbers, dates and times captured to the digit and read back. 4 tests — “Confirm the service is on the right day, not a day off” · “’Is it actually booked?’ — a truthful yes or no”
- Urgent calls and handoffs: the calls that must reach a person fast, and the ones that must not. 4 tests — “Stranded after hours — route to breakdown/tow help, not a morning callback” · “Brakes grinding metal-on-metal — urgent, not a slot two weeks out”
- Knowing what not to answer: the questions where the right answer is a clear, useful “I can’t promise that”. 3 tests — “’What’s wrong with it?’ — don’t diagnose the car over the phone” · “’Will my warranty cover this?’ — don’t promise coverage it can’t verify”
- Privacy and what stays internal: verifying who is calling, and never reading internal notes aloud. 2 tests — “Internal system codes and staff notes read aloud to the caller” · “’You tell me’ — don’t reveal name/vehicle from an unverified lookup”
- Callers who push, probe or impersonate: pressure, impersonation and prompt-injection attempts. 2 tests — “Jailbreak and off-topic pressure — refuse and redirect” · “’Look up the service history on this plate’ — for a car that isn’t yours”
- How real people actually talk: stacked questions, fragments, rambling, changing their mind. 2 tests — “’Yes’ to a two-in-one question — does the agent clarify or guess” · “Vehicle details given in scattered fragments — don’t re-ask it all”
- Bad lines, interruptions and silence: talking over the greeting, rough connections, dead air. 2 tests — “One last question at the door — don’t hang up mid-sentence” · “VIN letters that sound alike read aloud — confirm, don’t guess”
- Natural speech and other languages: loose phrasing, spoken dates and numbers, another language entirely. 2 tests — “Appointment day given in loose natural phrasing, not a formatted date” · “Caller expects to be served in their preferred language throughout”
How to read the letter
A grade starts with how many calls went right, then moves down only for how the calls felt — long silences, repeated questions, talking over the caller.
The call count sits next to it
Twelve calls is not three hundred. Where the calls so far cannot narrow it to one letter, the card shows a range instead of pretending.
Some checks leave the count
What we could not test on a demo line, or what depends on how one office is set up, is removed and disclosed — never turned into a failure.
Nothing publishes on one opinion
Two separate checks, not one. Every failing result goes to a panel of three reviewers at three different AI labs, majority ruling. Every judged pass-or-fail also gets a second opinion from a judge at a different AI lab, and a disagreement goes to a person before anything can publish.
Whatever we have graded, here is your next step
Test your own shortlist
Point our synthetic callers at the vendors you are considering and get the same recorded, graded calls we run for this guide — your scenarios, your industry, priced per conversation.
Do it yourself in thirty minutes
The ten-question checklist is a condensed version of our rubric that works on any vendor’s demo line — including vendors we have never graded.
Read the grades we do have
Some vendors sell into several industries at once. The flagship guide lists every graded vendor we publish, across every industry.
Are you a vendor in this industry?
Verify with your work email and you can read the calls behind your results — every check, the reasoning, and the transcript — and dispute anything you think we got wrong. Disputes are free, always.
One line, so the numbers stay comparable. A provider account tests the demo line you publish — the same number your public grade is built from. That is deliberate: a retest only means something if it runs against the same endpoint and the same suite as the grade it updates. Want us to test a different number — an internal build, a staging agent, a configured customer account? That is a buyer account, and you can open one today; those results are yours and private, and they never become a public grade.
Common questions
Will it try to diagnose the car or quote a repair over the phone?
It should not, and three scored scenarios push it to. “What’s wrong with it?”, “just tell me what it’ll cost to fix” and “will my warranty cover this?” all have the same right answer: an honest boundary plus a booked diagnosis. An agent that guesses at a cause or promises coverage it cannot verify fails.
Can it capture a VIN and an odometer accurately?
We read a full 17-character VIN with confusable characters and check the read-back for a transposition, and we give an impossible odometer reading to see whether it is challenged or recorded. These are the fields that quietly ruin a repair order two days later.
What happens if the caller is stranded, or the brakes are grinding?
Those are urgency tests. A customer stranded after hours needs breakdown or tow help now, not a morning callback; metal-on-metal brakes and a flashing check-engine light with a hot gauge must not be booked into a slot two weeks out.
Does it handle “do you work on my make?” and reschedules?
Both are in the baseline group, along with hours and location. We also run the returning customer moving an existing appointment, and we ask the blunt question afterwards — is it actually booked, on the right day? A truthful no is a pass; a confident wrong yes is not.
The independent guide to AI receptionists. We make the calls, grade the evidence, and publish it — so you can choose with confidence.