# First, About the Robot in the Bay — the opening of AI for Auto Repair Shops > First, About the Robot in the Bay — the opening of AI for Auto Repair Shops, free to read in full. 2,194 words on what AI does and does not do in this trade. *Free to read in full. 2,194 words, the first section of AI on the Job: Auto Repair Shops Edition.* You have been hearing about this for a while now, and nearly all of it turns up in one of two flavors. The first is the end of the trade. Cars are computers with wheels, the manufacturers own the software, and any morning now a machine will do the diagnosis, order the parts and hand a printout to a kid in a clean shirt. There is usually a video attached, and somebody’s nephew in tech who has seen the roadmap and is not allowed to say more. The second flavor arrives at the trade show wearing a lanyard. This thing is going to fill your bays, close every estimate, answer the phone and find you two techs by spring. Thirty days free, sign here. The two stories have the same shape. Both are enormous, both are about a year that has not happened yet, and neither leaves you a single thing to do when the lot fills up on Tuesday morning. Believe whichever one you like. The week does not move. The estimate you promised before lunch is still not written, the timing chain job is still sitting on approval, and the woman whose wheel bearing you did last month is still telling her neighbors you charged her for a noise that came back. This book has nothing to say about the future. It is about that estimate. Take the first story first, because it can be settled with a list, and the list is physical. It cannot road-test the car. It cannot get it out on the four-lane, bring it up to fifty-five and feel the shudder come in under light braking. It cannot sway a wheel at ten and two, put a bar under the tire and tell whether the play is the ball joint, the tie rod end or the bearing. It cannot hear a noise go quiet in a left-hand turn. It cannot smell coolant, and you can — from the doorway, before the hood is up, before anybody has said a word to you. It cannot hold a hand near a wheel and know which corner is dragging. It cannot back-probe a connector, watch live data while somebody loads the engine, or move a coil to the next cylinder and see whether the misfire follows it. It cannot get the seized bolt out. It has never snapped one off either, so it has no idea what that does to a Tuesday. It also does not sign anything. Your name is on the repair order. Your certifications are on the wall, your registration is on file with the state, and when the wheel goes back on it is your torque wrench and your comeback. No software carries any of that and none is coming that will. It does not know your state’s rules either, and this is the one worth being unfriendly about, because it will answer as though it does. Ask what has to appear on a written estimate where you are, or when an increase needs fresh authorization, and back comes a tidy confident paragraph describing some other state’s law, or nobody’s. Chapter 10 handles this properly. Until then, treat anything it says about the law as a rumor you overheard at the parts counter. Which is a long way of saying no. Not the wrench, not the diagnosis, not the judgment, not the paper that carries legal weight. That is the whole of the reassurance, and it fit on a page. None of which makes the worry silly, because it is not silly. Trades get rearranged all the time by software that never picks up a tool. Somebody slides an app between the person doing the work and the person paying for it, and the money quietly starts changing hands somewhere new. You already have the small version of that. The lead services. The booking platforms. The fleet and warranty networks that will rent you your own neighborhood by the car. The fact that a driver with a grinding noise types the symptom into a search box before she ever picks up a phone. So the thing actually worth watching is not a machine holding an impact gun. It is the question of who ends up standing between your bays and the person who needs brakes. Which argues for reading on. The shops that are hardest to get in front of are the ones that reply first, explain best and own their own reviews, and Chapters 7 and 9 are about nothing else. Now the part where a weaker book would tell you nothing is changing. Something is changing. You would catch the lie by Chapter 4 anyway, and then you would stop believing the useful parts. What is changing is not under the hood. It is the writing. Nobody warns you about that when you open a shop. You thought you were buying a business that fixes cars. What you bought was a writing business with a lift in the back. You write the estimate. Then you write the human version of the estimate, because “RF caliper seized, inner pad to the backing, rotor heat-checked” means nothing to the person who has to approve it. You write the approval request, then the follow-up to the approval request, then the parts-delay message. You write the answer to why the brake job is five hundred and change when the pads are forty-seven dollars online. You write the invoice line that stops the diag-fee argument before it starts at the counter. You write the reply to the review from the customer whose check-engine light came back on the way home. You write the declined-work follow-up — or, honestly, you do not, and next quarter’s revenue sits quietly in a stack of closed ROs. None of it is fixing cars. None of it is billable. Nearly all of it happens after the last car goes out, or not at all. That stack is the part these tools are genuinely good at. They cannot find the fault. They are very good at the six sentences that explain the fault to somebody who has been talked down to by a shop before and walked in braced for it to happen again. They are bad at things too, and nobody with a booth at the trade show is going to tell you which. The main one is that they make things up, in the same steady voice they use for the true parts. Ask for book time on a timing chain and you get a number that came off a forum post written by a man with a driveway and an opinion. Ask what a job goes for in your market and you get a figure invented on the spot. It does not know your labor rate, your parts matrix, or what your jobber charged you this morning. It has not seen the car. It has never set foot in your town. So every workflow in here hands it the sentences and keeps the numbers with you. Findings, measurements, prices, book time — yours. Wording — its. Where that line is easy to cross there is a WATCH OUT box with the check written out, and it takes about a minute. Call it the cost of using the thing. It is a good deal cheaper than the evening. And the commercial part, said once and plainly. Your customer cannot judge your wrench work from a waiting room. They judge what they can see: how fast the estimate came, whether anyone explained what a wheel bearing actually does, whether the shop called before the promised time or after. In a business where every customer walks in half-braced to be taken, the shop that explains gets the yes. None of that is new. The new part is the price of the explanation. It used to cost you an evening. It now costs about as long as a cup of coffee. There is a particular kind of Tuesday that anyone running three bays already knows. The lot was full by eight thirty. Three cars are diagnosed by eleven and all three need real money spent on them. One customer is a waiter — she is in the lobby with a laptop and has asked twice, politely, how much longer. The timing chain job from yesterday is still sitting on approval because the second signature is a spouse who has not called back. The phone has gone four times with “any update?” and every one landed while you were mid-sentence with somebody else. The first estimate goes out at two, and it is a good one, because you sat down and wrote it properly. Nine minutes on the phone afterwards and she says go ahead. The second one you do at four, faster and thinner, mostly numbers with eleven words beside them, and that customer says he wants to think about it — which is what people say when they do not understand what they are being asked to buy. The third one does not happen. It is six forty, the techs have gone and the counter needs clearing. You tell yourself first thing tomorrow. First thing tomorrow is a no-start in your own parking lot at ten past eight. On Thursday that customer books somewhere else and you never learn why. It was not price. The other shop texted him a written estimate with photos on Tuesday afternoon, while he was still thinking about it. Notice what the problem is not. You are not bad at explaining. You explained beautifully at two o’clock and the job sold on the spot. The problem is arithmetic: a good explanation takes forty minutes, and Tuesday handed you three of them and one of you. Here is what this book is, so you can decide now whether to carry on. It is a working manual for using an ordinary AI chat tool — ChatGPT, Claude, Gemini, take your pick — on the paper side of an independent shop. Estimates and inspection write-ups. Customer messages, approval requests, daily status texts, the nine o’clock “should I be worried” text. Invoices, warranty claims, fleet billing, the payment chase. Review replies. RO notes. Declined-work follow-up. The marketing you have been meaning to do since spring. There are 150 prompts, numbered, with a library at the back so you can find one in ten seconds with a customer at the counter. Every prompt has been run through ChatGPT, Claude and Gemini, and anything that only worked in one of them did not make the book. It is not a list of what these tools might one day do for a repair shop. It is the wording that came back usable. What it is not: a scan tool, a diagnostic aid, a replacement for your service information, a labor guide, or a way around your state’s repair statutes. It will not tell you what is wrong with the car, and nothing in here asks it to. Everywhere the tool gets close to something that can cost you — customer data, VINs and plates, authorization law, advertised prices, safety language — there is a RULES CORNER box marking the fence. Chapter 10 collects that fence into one chapter, and it is the one to read before a customer’s details go near a chat window. None of it needs you to be good with computers. It needs a phone, an email address and half an hour. Chapter 2 is the setup. Chapter 3 is the only real skill in the book and you have had it for years, because talking to one of these tools is writing an RO: complaint, what you found, what you want done. Brief it the way you would brief a tech you did not want coming back with three questions. A word about size, and then you can get on. Nothing in here changes your life. It changes what happens between five and seven, and after a month, most of your evenings. The Promise Page opposite puts a number on it, about eleven hours in a normal week, and that number is honest — but it is not free. The first week costs you time instead of giving it back, the way every new tool does. Somewhere in week three the sums start going your way. Two things out of this book and you are square with what you paid for it. Two things is what most people take: the estimate workflow and the daily status text, marketing chapter never opened, four or five hours back a week, and then they stop thinking about it. That is the right outcome, not a partial one. There are no prizes for finishing all twelve chapters. Nobody is coming for your bays. The car still goes up on a lift, and somebody with hands and judgment still has to find out why it is doing that. What is on offer is smaller and more useful than either story you have been sold: the estimate out the same day it was diagnosed, ten minutes instead of forty, and a Tuesday that finishes when the last car leaves. The next page shows you where the hours come from. --- The rest is twelve chapters, 150 numbered copy-paste prompts, a glossary and a thirty-day plan. - Book: ../books/001-ai-on-the-job-auto-repair-shops-edition.html - All 150 prompt titles: ../prompts/001-ai-on-the-job-auto-repair-shops-edition.html --- *AI on the Job Press · Greenlight Publishing · 2027 edition*