For years, supplement disputes were mostly a back-and-forth between the shop and the adjuster. The shop found the missing operation, explained why the repair required it, and backed the supplement with the relevant procedure or estimating documentation. That conversation still happens, but on a growing share of claims it's no longer where the review starts. An automated system has often already screened the estimate before an adjuster has looked closely at the repair. 

By the time the file reaches your desk, it may have run through a review that flagged, reduced, or stripped out operations according to rules the carrier configured. CCC Estimate-STP, Mitchell Intelligent Estimating, and Audatex's Qapter Intelligent Estimating now handle a meaningful share of U.S. auto claims. In early 2023, CCC reported 15 insurers using its STP program, seven of them in the top 10 by direct written premium, together representing about half of U.S. auto claims volume. For many shops, the first objection to an estimate now comes from a rules engine that ran the file before any person looked at what the vehicle needs. 

What the repair requires hasn't changed. What's changed is when the first pushback arrives, and who or what delivers it. 

These systems are sold on speed, consistency, and less human handling. The catch for repairers is that they can treat a required operation as an exception unless the estimate data supports it in the exact way the system is built to read. Your technician may know precisely why a calibration, a transfer operation, or a refinish step is necessary, but unless that need is documented in a form the review system recognizes, the reasoning never enters the file. 

Regulators have started to catch up. In December 2023, the National Association of Insurance Commissioners adopted a model bulletin on how insurers use artificial intelligence, and many states have since issued their own versions. The principle is straightforward. An automated claim decision is still a claim decision, and the carrier's still responsible for making sure AI-assisted outcomes comply with unfair claims settlement practices laws wherever it operates. The liability sits with the insurer, not the algorithm. 

Most shops already know when an estimate is light. You catch it after teardown, when the recycled part needs more work than expected, when the calibration target comes out, or when the refinish calls for steps the estimate never paid for. What's harder to see is how far upstream the omission started, and how much stronger a supplement gets when you treat the dispute as a documentation problem instead of a negotiation. 

The four areas below are where these cuts show up most consistently. For each, here's why it happens and what documentation actually moves the file. 

ADAS Calibrations 

Calibrations now appear on 28.3% of repairable estimates, up from 21.8% the year before, roughly a 30% jump in a single year (CCC Crash Course 2026). The work is even more common than the estimates show. Removing or replacing a sensor, camera — or the panel it mounts to — is what triggers the calibration, yet the line that should follow often isn't there. When the estimate doesn't spell out a trigger that the system can read, automated review flags the calibration as duplicative or unsupported, making a coverage call from what's on the page rather than from what the technician found on the car. 

A supplement that only says a calibration is needed gives the reviewer room to say no. Show why instead. Attach the current OEM procedure, the platform documentation confirming that scan and calibration time isn't included in base labor, and the relevant DEG inquiry. With those in the file, rejecting the line means rejecting the OEM requirement, the estimating database, and the published estimating clarification all at once, in writing. 

Refinish on Adjacent Panels 

Blend time is the clearest example of the gap between what the data supports and what the default pays. All three estimating systems, CCC, Audatex, and Mitchell, calculate blend at 50% of a full refinish. The SCRS blend study found close to the opposite. Blending takes 31.59% more time than a full refinish once you account for the work needed to match color and texture. The information providers responded by letting estimators override the default, but 50% is still what the systems assume out of the box. 

So when you raise blend time to reflect the actual work, you're correcting a default the study contradicts, and automated review reads the increase as a deviation and flags it. The same happens with protective coating removal, de-nib and polish, all treated as included when the platform language confirms they aren't. None of it is padding, just payment for the time the finish actually takes. 

Build the response around the SCRS study and the platform's own not-included language and attach both. Point to the specific page or database entry instead of trading opinions. Rather than asking the reviewer to accept your number, you're showing that the published research and the platform's own documentation back the operation, and the only thing arguing against it is a default formula that the industry's own research has already called into question. 

Parts Downgrades 

When an automated sourcing query swaps an OEM part for a recycled or aftermarket one, the labor time usually stays the same, and that's the problem. Base labor times are built around a new OEM part that arrives ready to install. A recycled assembly doesn't arrive that way, and it may require transferring brackets, clips, fasteners, wiring, moldings, and harness routing, work the new part never needed and the original time was never written to cover. The system made a parts decision without making the matching labor decision, and that missing labor is the supplement. 

A common version is an OEM door shell swapped for a recycled assembly while the transfer labor drops off or never gets added. The part itself may be perfectly good, but the missing labor is a separate question. 

Cite the platform's included and not-included operations for recycled and aftermarket parts. When the part is structural or safety-related, the OEM position statement carries more weight than any platform page, because nearly all OEMs have issued guidance limiting or prohibiting certain recycled or non-OEM structural and safety-related parts and directing you to their specific repair procedures. At that point, the dispute isn't really about cost. It's about whether the repair can be completed safely under the manufacturer's published procedures. 

P-Pages and Included Operations 

P-page disputes have shifted because the source material is no longer static. MOTOR, which develops the database CCC runs on, retired the standalone PDF of its Guide to Estimating, and the current version lives inside the platform. A binder or saved PDF a shop has relied on for years may no longer match it, so a reviewer working from a newer reference and a shop working from an older printout can both cite the P-pages and still be reading different material. 

The mismatch shows up as missing operations. A panel replacement comes through without glass, trim, or transfer operations. Refinish gets written with no masking allowance beyond the panel perimeter. An EV repair shows up with no high-voltage isolation procedure. These often aren't arguments about whether the operation belongs. In many cases, the operation is simply absent, sometimes because the reference being used is out of date or incomplete. 

Before writing the supplement, pull the current platform documentation for each flagged operation and attach the page that shows whether it's included or excluded. With that attached, saying no means disagreeing, in writing, with the published rules of the system they're already using. Most reviewers aren't willing to do that. 

Closing 

The repair standard itself is the same as it's ever been. The vehicle needs what it needs, and the OEM procedures, estimating databases, and industry documentation spell that out more clearly than ever. What's shifted is how early the first decision gets made, and how much of it happens before a person has even engaged with the repair. 

The shops handling this well don't rebuild the argument after a denial. They put the documentation in the file from the start. The calibration trigger, the OEM procedure, the platform methodology page, the current P-page reference, the DEG inquiry, the not-included operation language, all of it attached before the review runs. With all of that already in place, a reviewer has little room to treat a required operation as your preference, and flagging the line becomes harder than approving it. 

There's a regulatory layer worth watching, too. State insurance regulators now have clearer authority to examine how carriers use AI in claims handling, including how those systems are validated and whether their outputs comply with unfair claims settlement practices laws. A documented pattern of reductions on required operations that the platform and the OEM both support eventually stops being a supplement problem and becomes a compliance one. 

That doesn't mean every disputed line becomes a Department of Insurance complaint, but it does mean you're right to track patterns. When the same carrier keeps removing the same documented operations despite OEM procedures, database support, and your supplement documentation, the problem has outgrown any single estimate, and it's worth knowing where your state stands. The NAIC bulletin is guidance, not law, and it carries weight only where your Department of Insurance has adopted it. Adoption is uneven, so check your DOI website or ask directly whether they've adopted it and what it requires of carriers in your state. That tells you how much the compliance angle is actually worth in your market. 

It all comes back to transparency. A carrier using AI to screen estimates still has to stand behind what the system produces, and attaching the OEM procedure, the database page, and the DEG inquiry forces that into the open, where the reviewer either accepts what the published sources support or rejects them in writing. That's the accountability the rules are reaching for, and it gets built one clean file at a time.