AI automation for body shops: score higher

Your panel work is solid. Your paint shop matches colour on colour. And still your score with the insurer slips, and you can't quite say why. That's because you aren't judged on how well you repair a car, but on a quarterly scorecard where communication, turnaround time and admin count just as heavily as your craftsmanship. So AI automation in a body shop is almost never about the car, it's about the points you're losing while you're welding. In this article I explain honestly which points you can actually win, and which ones you should leave well alone.

Your flow of work is less and less yours

First the state of play, because it drives everything. Consultancy KPMG carried out a network study for BOVAG Schadeherstelbedrijven, the Dutch body shop trade association, looking ahead to 2030. It found that the share of steered repair work is growing from around 60 percent today to between 67 and 82 percent by 2030. Translated to your workshop: soon up to four in five cars arrive because an insurer or leasing company sent them to you. Not because the customer saw your sign by the road.

At the same time the number of repairs is falling slightly, around 0.3 percent a year, mainly because driver assistance systems prevent collisions. But the average repair bill rises, according to the same study, from just over €2,200 to around €3,400 by 2030, because cars get more complex and parts get pricier. So there's plenty of work, as BOVAG itself points out. The catch: who gets that work is increasingly decided by somebody else. And KPMG warns that pre-tax profit per site could fall by around 2.5 percentage points by 2030 if nothing changes.

A scorecard decides how much work you get

If you're in the Schadegarant network, you know eXperate. Every quarter you get new scores, and they help determine your position. Schadegarant measures four main objectives, and they're set out plainly in its own explanation of the system:

• Quality and safety, judged by the policyholder, plus physical and sample-based audits of your files and parts
• Customer satisfaction, worth up to 20 points, built entirely from what your customer thinks and what you record
• Process efficiency, do you submit your file on time, are your photos right, is your estimate right
• Repair cost, up to 30 points, based on how your hours, parts and paint materials deviate from the average of every other repairer

Look at that closely. Of those four, exactly one is about how you repair a car. The other three are about how you communicate, what you record and when you submit it. That's not a criticism of the system, it's simply where your points are. And it's exactly the part nobody in most body shops has time for, because everyone is out in the workshop.

The points you can win without touching your craft

Take customer satisfaction alone: 20 points. They're built from a customer survey (10 points, covering reception, the time the repair took, the information you gave, the explanation at handover and the replacement vehicle), the Net Promoter Score (7 points), email addresses supplied (1 point) and complaints (2 points). Four of those five survey questions are about talking to your customer. Not about welding.

And watch the thresholds, because they're hard. A survey mark of 7.0 or lower earns you zero points, not half a point, zero. Complaints work the same way: stay under 1 percent of upheld complaints and you take the maximum; hit 1 percent or more and you get nothing. For email addresses, the average across the whole collective is 73 percent: above 70 percent you take the point, below 45 percent you get nothing. And in process efficiency, for both your photo protocol and your estimate: below 70 percent approved is zero points.

Where AI does make a difference in a body shop

No theory, just the work that's left undone because your technicians aren't receptionists:

• Keeping the customer posted, 'your car is in the paint shop, we expect it Friday' goes out automatically the moment your planning shifts. That hits the survey questions on turnaround and information directly.
• Capturing and checking the email address at intake, so that point no longer sits empty in your file by accident.
• Flagging that a file has to be submitted before 4pm on the day the repair starts, before it's too late, not after.
• Checking the photo set is complete per protocol before the car leaves the workshop and nobody has it on the lift any more.
• Answering the phone when nobody can. Every missed caller is a customer wondering where their car is, and that comes back in your score.
• Catching a complaint the moment it arrives, rather than once it's been upheld. Staying under 1 percent is easier than reversing a complaint.

The lookup work nobody puts on a timesheet

There's one more thing, and the good part is the industry has already written it down. ABS Autoherstel talks openly about its AI approach, and there Edward Reinderts names a figure you'll probably recognise: 20 to 30 minutes are lost per file looking up agreements in manuals, Excel files and PDFs. Because every client applies its own terms. Can you repair a bumper for this insurer or must it be replaced? Which discount percentage applies here? When may you provide a replacement vehicle?

ABS is honest enough to add that standardising those agreements is unworkable in practice, precisely because every client has its own rules. That's an important conclusion. It means you'll never solve this with an off-the-shelf package that works the same for everyone. Your terms belong to your clients.

What a chain builds, and what you can build

ABS solved that with Herbie, an in-house assistant that answers their front office within seconds about client agreements. They're also running a trial with automatic photo-based estimating, and working on smarter planning. Nicely done. But that's what a chain with dozens of sites can put together.

If you're fourteen people in one building, you read a story like that and think: good for ABS. And that's exactly the misunderstanding I want to clear up. You don't need a chain's budget to make your own terms searchable, or to keep your customers posted automatically. You don't need a trial across five sites to get an email address into your file. These are bounded jobs of a few weeks, not a few years. The difference between you and the chain isn't the technology, it's that they have somebody whose job it is to sort it out, and you don't.

A quick word on the rules

If you let an assistant talk to your customers, from 2 August 2026 you have to make clear they're dealing with AI. That's the new European AI law. Sounds annoying, but in practice it amounts to little: one line in your message and you're done. I build it that way as standard, so it's not something you need to think about. What I mainly want you to take away: this is no reason not to start. It's one sentence.

What does this cost?

I'm transparent about it, these are my prices. For a body shop it usually looks like this:

• Getting acquainted (30 min · free): Where does your time go, and which systems do you use? We discuss how your business works.
• Small project (€300–€5,000, 1 to 2 weeks), for example keeping your customers posted on their car's status automatically, or making your client terms searchable.
• Medium project (€5,000–€12,000, 3 to 6 weeks), several things joined up: intake, status updates, file flagging and your phone.
• Maintenance (€100–€750 a month for basic, €750–€1,500 for standard), because your clients change their terms, and it has to move with them.

A flat monthly fee, not a rate per message or per file. That's a deliberate choice: you know where you stand up front, and it doesn't get more expensive the better it works.

When not to start with this

I don't sell AI for the sake of AI. Skip this if:

• You take walk-in work and aren't in a steered network. Then there's no scorecard and you earn your money another way, this story isn't for you.
• You run a handful of files a week. Then just call those customers yourself, and it's more enjoyable anyway.
• Your scorecard is already high and you know why. Don't fix what isn't broken.
• What you really have is a capacity problem in the workshop. No assistant ever sprayed a bumper. If you're short of technicians, solve that first.
• You're about to be acquired or to switch systems. Wait it out, or you're building on sand.

How to start

• Pull out your last eXperate scorecard and see where you're leaving points. Not your total, the individual components.
• Cross off the repair-cost points. You solve those with craftsmanship and with your client, not with software.
• Count what's left. Usually it's things like email addresses, turnaround, information given, files on time.
• Ask your front desk how much time they lose looking up what a client does and doesn't allow. Convert it to hours a week.
• Start with the point that pays most and touches your craft least. Usually that's keeping the customer posted.