AI Insights Automotive
Fred: A Service-Desk Assistant That Won’t Diagnose Sight-Unseen
Talk to Fred
Ask Fred about Automotive
This is the same Fred you would put on your own site. Ask about Automotive, compliance, or how the guardrails work. Fred listens.
A driver messages your shop’s site about a grinding noise when they brake, asks what is wrong, what it will cost, and whether it is safe to drive until Friday. A general-purpose chatbot diagnoses worn pads, quotes a price, and says it is fine to drive. No one put the car on a lift. The shop wanted a tool to book more inspections. It got one diagnosing, quoting, and clearing a vehicle for the road that nobody had seen.
The threat and standard articles in this series walk through how that turns into a bait-quote complaint or worse. This one is the answer: what a compliant AI assistant for auto repair looks like, and why the design, not a disclaimer, keeps a website from making calls only a technician can make.
The Real Choice Is Governed or Ungoverned
Shops ask whether to use AI on the site at all. Drivers already decided. They want to describe a problem, ask about hours, and book a visit without waiting for business hours, and the shop that makes them wait loses the work. The real question is whether the assistant is governed.
An ungoverned chatbot names the problem, quotes the fix, and clears the car to drive. A governed one books the inspection, explains how an estimate works, and routes diagnosis, pricing, and safety calls to a technician. Same convenience, completely different exposure when the inspected estimate and the website disagree.
How Fred Holds the Line
Fred starts from the opposite default of a general chatbot. A generic bot answers anything unless told not to. Fred answers only what it is cleared to, and the boundary lives in the system rather than a prompt that drifts when a customer pushes.
In practice, Fred works from your own content. It books inspections, explains your services and hours, and captures the symptom and vehicle details the technician needs. It does not diagnose the problem or quote the repair sight-unseen, because most state repair acts assume a written, authorized estimate after a real inspection, and a blind number invites a bait-quote claim. It does not say a car is safe to drive, and it does not promise a repair is "guaranteed" or "covered," because a written warranty representation can pull in the Magnuson-Moss Warranty Act and a misstatement sits within the FTC’s deceptive-practices authority. Fred runs more than 50 industry guardrail packs, and the auto repair pack is built around the line between booking a job and diagnosing one.
The design proves itself under pressure. A driver who wants a number will rephrase "what’s this going to cost" several ways, and a prompt-instructed bot eventually answers the version it was not warned about. Fred does not rely on catching the phrasing. It decides what may be said before the reply forms, so the blind quote and the safety clearance never reach the customer. "Will not" is a suggestion. "Cannot" is an architecture.
Fred vs. a Generic AI Chatbot
| Situation | Generic AI Chatbot | Fred |
|---|---|---|
| "What’s wrong with my car?" | Diagnoses from a description | Books the inspection; a technician diagnoses |
| "What will the repair cost?" | Quotes a number sight-unseen | Routes to a written, authorized estimate |
| "Is it safe to drive this week?" | Clears the car for the road | Declines the safety call; books the visit |
| "Is the repair guaranteed?" | Promises coverage | Leaves warranty terms to the shop |
| Where the rules live | In a prompt a customer can push past | Built into the system; enforced before output |
| Who owns the regulated answer | Effectively the bot, and your shop | The technician who has seen the car |
That single screen is the argument. A general chatbot is helpful until helpful becomes a diagnosis, a bait quote, or a safety clearance. Fred is helpful across everything that moves a job forward and structurally incapable of making the call a technician owns.
What Your Shop Actually Gets
Set the rules aside and look at the phone. Fred handles the after-hours questions that usually die in voicemail, are you open Saturday, do you work on my make, how soon can I get in, what should I bring, and answers them instantly, overnight, in plain language. It captures the symptom and the vehicle so the write-up starts warm. Everything that commits a diagnosis, a price, a safety call, or a warranty stays with a technician.
The captured-lead math is the real win. The driver who messages at 9 p.m. and gets a useful, accurate reply is the one who books, instead of the one who called the next shop while waiting for a callback. Fred turns the after-hours window into a pipeline without ever quoting a repair it cannot stand behind.
So the question is not whether your competitors will put AI on their sites. They will. It is whether yours is governed before a driver holds it to a number, or a safety call, it never should have made.
Frequently asked questions
Can a compliant assistant give any kind of estimate?
It can explain how your estimates work, what an inspection involves, and what factors affect a repair, the same context a good service advisor gives. What it will not do is quote a repair on a car it cannot see, because most state repair acts expect a written, authorized estimate after an inspection and a blind number invites a bait-quote complaint. Fred books the inspection and routes pricing to a technician.
Why won't it say whether a car is safe to drive?
Because that is a safety judgment about a vehicle no one examined, often on a symptom that can signal a real hazard. If a customer relies on it and something fails, the shop has made a documented safety representation with nothing behind it. Fred declines that call and books the visit so a technician who has seen the car can make it.
How is Fred different from a generic chatbot told not to diagnose?
A prompt instruction is something the model abandons when a driver rephrases "what’s wrong and what’s it cost" in a way it did not anticipate. Fred enforces its boundaries at the system level, deciding what is allowed before it responds, so a diagnosis, a blind quote, or a safety clearance is never generated regardless of wording. That is the difference between a bot that usually deflects and one that cannot make the call.