Count the Supplements You Wrote in June. That One Number Tells You If AI Blueprinting Paid for Itself.

How many supplements did your shop write in June, and how many of them were things a good blueprinter could have seen on the car the day it was torn down?

Most owners cannot answer the second half without pulling files, and that is the point. Supplements are so normal in collision work that they stopped being treated as a defect. They are treated as weather. A car comes in, gets written, gets torn down, gets written again, and then again when the part shows up wrong, and each round trip to the insurer costs you days of calendar time on a car that is sitting in a stall doing nothing.

Deloitte's State of AI in the Enterprise 2026 found that 84 percent of organisations investing in AI report positive returns, and the pattern behind the ones that do is boring: pick one messy process, keep a human reviewing the output, and measure whether time went down or errors went down before letting it anywhere else. For a body shop, the messy process with the biggest number attached is the supplement.

What is your blueprint bay actually missing?

A blueprint check is a second read of the disassembled car: the estimate you wrote, the photos your blueprinter took, the pre-repair scan report and the manufacturer's procedure for that exact VIN, all compared against each other before the car leaves the teardown stall. Done properly it is the single most valuable half hour in the building. Done at 4pm on a Friday with two more cars waiting, it becomes a walk-around and a hunch.

The misses are predictable, which is what makes them measurable. One-time-use fasteners and bolts nobody wrote. Corrosion protection, seam sealer and weld-through primer left off because they are not-included operations and the estimator was moving fast. A calibration requirement on a car with a forward camera that the pre-scan flagged but nobody carried onto the sheet. A blend into an adjacent panel that gets discovered when the painter opens the can. Bracket and absorber parts hidden behind a cover. Each one on its own is a $60 to $400 line. Together they are the reason you are writing 1.9 supplements a car instead of 1.1.

What should the assistant do, and what stays with the estimator?

The job you are handing over is reading and cross-referencing, not deciding. It reads the estimate you already wrote in CCC ONE, Mitchell or Audatex. It reads the teardown photos. It reads the pre-repair scan report from whichever tool your shop uses. It reads the manufacturer's repair procedure for that VIN and any position statements that apply. Then it hands your estimator a draft list of lines that are missing, each one with the reference it came from attached.

Your estimator approves, edits or deletes every line. Nothing goes to the insurer that a licensed human did not sign off. That human review step is not a nicety — it is the thing that separates the 84 percent who report a return from the shops that switched something on and quietly turned it off in six weeks.

flowchart TD
  A["Car in the blueprint stall, 100 percent teardown"] --> B["Blueprinter photographs damage and runs the pre-repair scan"]
  B --> C["Assistant reads estimate, photos, scan report and OEM procedure"]
  C --> D["Draft missing lines, each with its procedure reference"]
  D --> E["Estimator approves, edits or deletes every line"]
  E --> F["One supplement uploaded to the insurer"]
  F --> G["Logged: supplements per RO"]
  F --> H["Logged: days the car waited on approval"]

Which numbers do you write down before you switch anything on?

This is the step almost everybody skips, and skipping it is why the argument about whether it worked never gets settled. Before you change one thing in the blueprint bay, pull the last 90 days out of your estimating system and write down five numbers:

• Supplements per repair order, counted, excluding total losses.
• Average dollars per supplement.
• Days from supplement uploaded to insurer approval, split out by insurer, because they are not the same.
• Keys-to-keys cycle time, and separately, days a car sat with no touch time while waiting on approval.
• Estimator minutes per supplement. Time three of them with a stopwatch if you have never measured it.

The number that settles the argument is supplements per repair order. Everything else moves for a dozen reasons. That one moves when your blueprint gets better, and it is sitting in your system already.

What the arithmetic looks like on a 90-car month

Illustration, not a measured result. Substitute your own figures.

The estimator hours are the easy half of the case: 36 hours a month is most of a week of somebody who should be selling work and talking to customers instead of retyping lines. The cycle time is the half that actually moves your business. Taking 1.1 days off an 11.4-day keys-to-keys average is roughly ten percent more throughput out of the same stalls, the same painter and the same booth — and on a DRP scorecard, cycle time and rental days are the columns you get graded on hardest.

When does the number lie to you?

Supplements per RO is honest only if you compare like with like. Three things will make it lie:

A hail event. If a storm rolls through in April and you take 200 CAT cars, your severity mix collapses, your supplement rate drops on its own and you will credit the wrong thing. Exclude CAT work from both sides of the comparison, or compare the same months a year apart.

A change in mix. More heavy hits means more supplements no matter how good your blueprint is. Band your comparison by severity — under $2,500, $2,500 to $6,000, above $6,000 — and compare inside the bands.

A new insurer or a lost one. Adding a network changes both approval behaviour and cycle time. If your DRP roster changed during the trial, the trial is contaminated. Wait, or exclude that carrier.

And one honest limitation: a supplement you cannot avoid is not a failure. Hidden damage behind a rail is hidden damage. The target is not zero. The target is the ones that were visible on the car at teardown and got missed anyway.

What stays human

The phone call with the adjuster stays human. So does the decision to push on a not-included operation with one carrier and let it go with another, because that is a relationship judgement no software has any business making. The total-loss call stays human, and so does the conversation with the customer when the car they wanted back Friday is going to be next Wednesday.

There is also a quiet risk worth naming. A draft list of lines is very persuasive at 4pm. If your estimators start approving all of it without reading, you have not improved your blueprint — you have automated over-writing, and the first audit from a carrier will find it. Spot-check ten sheets a week against the draft and look at what got deleted. If nothing is ever deleted, somebody has stopped reading.

What to do on Monday

Pull the 90-day baseline. That is the whole first step and it costs you an afternoon in the office with your estimating system's reports. Then run the new check in parallel on ten cars without changing anything you send to the insurer — write the sheet the way you always do, and separately note what the draft list would have added. Compare. If the draft finds three or four real lines a car that your team missed, you have your answer before you have spent anything meaningful, and you have it in your own numbers rather than somebody's brochure.

Frequently asked questions

My DRP scorecard already tracks cycle time. Is that not enough proof?

Cycle time moves for parts delays, tech absence, paint booth downtime and the weather. It is a good outcome measure and a terrible proof measure. Supplements per repair order is much closer to the thing you actually changed, which is why it settles arguments and cycle time does not.

We do 25 cars a month. Is that enough data to prove anything?

Not in 30 days. At 25 cars you need a full quarter on each side to see past the noise, so plan on six months start to finish. The upside is that a small shop feels the estimator hours immediately, because in a shop that size the estimator is usually also the person answering the phone and moving cars.

What if the insurer just denies the extra lines anyway?

Some will. Denied is still better than missed, because a denied line with a procedure reference attached is a negotiation you can have once, at blueprint, rather than a discovery three days into the repair that stops the car. Track your approval rate on the added lines by carrier — after a quarter you will know exactly which arguments are worth making.

Do we have to buy new estimating software?

No. Whatever you write in today is what it reads. The change is in the blueprint stall and in what the estimator sees before uploading, not in your estimating system.

Every day a car sits waiting on an approval is another round of customers calling to ask what is happening, and those calls land on the same two people who are trying to write the next sheet. CallSphere builds AI voice and chat agents that pick up the shop line and website chat outside of counter hours, capture the caller, the vehicle and the claim, book drop-offs, and route anything that needs a person straight to one. It will not write your supplement — but it stops the front counter from paying for the wait twice.