AI software for auto shops: eight questions to ask before you buy

Before you sign anything, ask the vendor eight questions about your own shop. Which system does it read? Can it write back? Who approves what it sends? What happens when the feed is down? A good vendor answers each one in a sentence, without slides. If the answers are vague, the product probably is too.

We sell AI software to auto shops, so read this with that in mind. The questions work for any vendor, including us.

Start with the job, not the product

"AI software for auto shops" covers very different products. A phone agent, a review-reply tool, a declined-work follow-up tool and a multi-shop reporting layer have little in common besides the label. Write down the one job you want done, the person who does it today, and how you'd know it got better. That sentence is your test for every demo.

Good jobs to start with are ones that are repetitive, have a clear owner and leave a trail in your shop-management system: open estimates, unanswered reviews, callbacks owed, parts delays.

The eight questions to ask any vendor

1. Which system does it read, and how?

Shop-management systems decide who gets in. Tekmetric, for example, lists 70+ partner integrations and handles API access through an application process. Ask whether the vendor has an approved connection to your system, a workaround such as exports, or only a plan. Ask to see your own data in the demo, not a sample shop.

2. Can it write back, and what exactly can it change?

Reading an estimate is low risk. Changing a repair order, a price or an appointment is not. Get a list of every field it can write, and who switched that on.

3. Who approves what it sends to customers?

Look for a named person per action type: the service advisor for estimate follow-ups, the manager for review replies, the owner for refunds. If the answer is "the AI handles it", ask what happens when it promises the wrong price.

4. What does it do when data is missing or a feed is down?

This is the question vendors least expect. A safe system shows the gap and waits. An unsafe one fills it in with a guess. The NIST AI Risk Management Framework, a voluntary federal guide released in January 2023, organizes this kind of thinking into four functions: govern, map, measure and manage. You don't need to read it to ask the question, but serious vendors will know it.

5. What record is left afterward?

You want to see the original data, the draft, who approved it, and when. Without that record, nobody can sort out a customer complaint or a wrong charge later.

6. Does it call or text customers, and is that legal in your state?

Voice tools need care. In February 2024 the FCC ruled unanimously that AI-generated voices count as "artificial" under the Telephone Consumer Protection Act, the law that limits robocalls. Answering inbound calls is a different situation from placing outbound ones, but ask the vendor how consent is handled for every outbound call and text.

7. Can they show you real results, and are they labeled honestly?

Ask for a customer you can call. Representative or demo numbers are fine if they are labeled. Unlabeled ones are a red flag, and regulators agree: in 2024 the FTC brought five cases in one sweep over deceptive AI claims, with the chair saying there is no AI exemption from the laws on the books.

8. What does it cost, all in?

Per-location fees, per-message fees, setup work and connection costs add up. Get the number for your shop count and your systems, in writing, before the pilot.

Why communication tools come up first

Most shops start with customer follow-up because that is where the complaints come from. In Cox Automotive's 2025 survey of 1,974 vehicle owners, 45% were unhappy with dealership service, mostly over unexpected costs and poor communication, and 55% said comparing costs online was very important. Independent shops are picking up those customers. The shops that keep them are the ones that call back when they said they would.

That is a good first workflow for AI software: it is measurable, it already has an owner (the advisor), and the risk of a draft that a person reviews is low.

A simple scorecard

Question	Good answer	Warning sign
Reads which system	Names your system and the connection method	"We integrate with everything"
Writes back	A short list of fields, off by default	Full write access from day one
Approvals	A named role per action	"Fully autonomous"
Missing data	Shows the gap and waits	Fills in a best guess
Record	Draft, approver and time, kept	Only the final message
Outbound calls and texts	Explains consent handling	No answer
Proof	A reference you can call, or labeled examples	Big numbers, no source
Price	Written quote for your setup	"Depends" with no follow-up

Where RevvOS fits, and where it doesn't

RevvOS is built for shops and groups that already run a shop-management system and lose track of follow-through between systems. It reads from your systems, flags what needs attention, and prepares a draft for the person who owns it. Customer and financial actions start as drafts by default, which is laid out on our security and authority page.

It is not a shop-management system, a phone system or a diagnostic tool. If you need help diagnosing a car, look at tools built for technicians. If you have one shop, one advisor and a short list of open estimates, a disciplined daily routine in your current software may be enough for now.

Our integrations page lists which connections are built and which are only represented for planning. We do not post a price; we scope a 45-day pilot on one workflow after an assessment, as explained on the pilot scope page.

Next step

Take the scorecard into your next demo, ours or anyone's. If you want to run it against RevvOS, book an assessment and bring the one workflow you'd fix first. For the bigger picture of what these tools do, see automotive AI software for repair shops.