What service writers actually want from AI — and what they don't.

Every week, someone tries to sell an auto shop new AI software. A new AI phone agent, a new AI estimate writer, a new AI "intelligent dashboard." The pitches are polished. The demos are clean. And the service writers I talk to end up using almost none of it.

I wanted to understand why. Over the past few months, I've had dozens of conversations with working service writers — mine, friends at other shops, folks in franchise networks across the Northeast. Some run front counters at Meineke locations. Some work at independents. Some are five years in, some are twenty-five.

What I heard was surprisingly consistent. Service writers aren't anti-AI. They're anti-bad-AI. And the difference between what they'd actually use and what's being sold to them is bigger than most vendors realize.

Here's what came through loud and clear.

What they want.

Something that drafts the annoying message so they can send it in 10 seconds.

The number one thing, across every conversation: "give me drafts, not decisions." Service writers spend real chunks of their day writing follow-up texts, status update emails, apologies, declined-work nudges. Not hard work, just time they don't have.

What they want: an AI that reads the RO, drafts the message, and drops it into a place where the SW can glance, edit a word or two, and send. They don't want the AI to send it on its own. They don't want the AI to decide tone for them. They want the typing done so they can focus on the person standing in front of them.

Something that remembers the customer so they don't have to.

A good service writer is part mechanic, part therapist, part memory palace. They remember that Mrs. Johnson's daughter just had a baby, that the Silverado out back belongs to a customer whose father passed last spring, that the guy in the waiting room always brings donuts.

They can't actually remember all of it. Nobody can. What they want from AI isn't a replacement for that intuition — it's a safety net. A two-line summary when the customer calls in: "Last visit 4 months ago, declined rear brakes, daughter drives the car, prefers texts over calls." Something that gives them the right instinct before they pick up the phone.

Something that catches mistakes before they go out the door.

This one surprised me. Multiple SWs independently mentioned: they'd love an AI that reviews the RO before it gets printed or texted to the customer. A quiet second set of eyes. Did you forget to add the alignment? Is the labor time obviously off? Did you miss the tire pressure sensor that was on the diagnostic?

Not a workflow replacement. A review layer. Something that sits next to the SW and says "hey, quick check on this one" before it gets sent. The word that came up over and over was backstop.

What they don't want.

AI that talks to customers on its own.

I have not yet met a working service writer who is excited about AI phone agents. Not one. Not because SWs are afraid of losing their jobs — most of them would happily never answer a phone again — but because they've watched what happens when automated systems get loose with customers.

The fear is specific: a phone agent that promises something the shop can't deliver. A chatbot that quotes a price that turns out to be wrong. A voice assistant that books an appointment for a service the shop doesn't offer. And then the customer shows up, and the service writer has to clean up a mess they didn't make.

The worst thing an AI can do to a service writer is lie to their customer.

AI that generates content for them to send.

There's a subtle but important distinction between drafting a message based on real facts and generating a message from thin air. Service writers can spot the difference in two seconds, and so can customers.

Nobody wants to send a text that says "Dear valued customer, we hope you're having a wonderful day! It's been some time since your last visit and we'd love to see you again!" It's hollow. It's obvious. It's insulting to both the SW and the person receiving it. AI that produces this stuff doesn't save time — it creates embarrassment.

Dashboards. All the dashboards.

Every AI vendor wants to sell shops an "intelligent dashboard." Service writers, mostly, don't look at them. They have two screens already. They have reports they can already pull. What they don't have is the time to turn data into action.

If the dashboard just shows them numbers, it's furniture. If it tells them "you have 7 declined jobs over $500 from the last 90 days that haven't been followed up on, and here are the drafted messages" — now it's useful. The difference is whether the AI does the next step, or just puts the information in a prettier box.

The pattern.

If you hold all of this up at once, a pattern emerges. The things service writers want from AI are the things that reduce toil without adding risk. The things they reject are the things that add risk in exchange for efficiency.

A follow-up draft they can review in 10 seconds: reduces toil, no risk. A phone agent that talks to customers solo: possibly reduces toil, but adds the risk of a miscommunication they'll have to fix anyway. Service writers, being good at their jobs, have done the math. They know which trade is worth it.

The vendors selling them AI tools have often not done this math. They're optimizing for impressive demos, not for the reality of a Tuesday afternoon with three customers in the lobby and a phone ringing. The demo shows you an AI that can handle everything. The reality needs an AI that can handle one thing really well and stay out of the way.

What I'm building because of this.

Most of what I work on these days is shaped directly by these conversations. The declined work follow-up tool I'm piloting right now? It drafts messages, routes them through the SW for review, only sends on explicit approval. The vendor receipt automation? It files the paperwork and produces a link — the SW still decides what goes on the RO note.

It's not as demo-impressive as "fully automated AI service advisor." It's better in every way that matters.

If you're a shop owner looking at AI tools, the test I'd offer is this: ask your service writers what they think. Not in a meeting. Casually, at the counter. Show them the demo. Watch their face. If they ask five sharp questions about use cases, buy it. If they say "that's cool I guess," don't.

They know. They've been watching tools come and go for twenty years. They'll tell you.