When an adjuster tells you what your diminished value claim is worth, there is a reasonable chance no human calculated that number. AI insurance claim decisions have moved from pilot projects into routine claims handling, and the shift happened faster than most drivers in North Carolina and South Carolina realize. What changed in 2026 is that regulators finally started building a framework to look inside those systems.
That matters to you for a specific reason. If a machine produced the number on your claim, the insurer still owns the outcome. Regulators have made that point clearly, and it gives you a line of questioning that most claimants never think to use.
What Insurers Are Actually Using AI For in Claims
This is not speculation. In its own survey work on property and casualty insurers, the National Association of Insurance Commissioners documented where AI sits in claims operations. Three uses came up consistently:
- Accident image analysis. Software reads the damage photos and classifies severity.
- Estimating ultimate claim settlement values. Models project what the claim should cost the insurer.
- Fraud detection. Systems score claims for anomalies and flag files for further review.
Read that list again with a diminished value claim in mind. Damage severity from photographs and projected settlement value are the two inputs that decide what a diminished value claim is worth. Both are now commonly handled by software before an adjuster ever picks up the file.
Regulators Started Building the Review Framework This Year
In 2026 the NAIC launched a multistate pilot of what it calls the AI Systems Evaluation Tool. It is a structured questionnaire that state examiners use during market conduct exams to look at how an insurer actually deploys AI. The pilot runs through September 2026, the tool gets revised in the fall, and adoption is expected at the NAIC Fall National Meeting in November 2026.
The framework asks insurers four things: where AI is used across the company, how it is governed, which systems are high risk, and what data feeds them. That last one matters most to claimants, because a valuation model is only as good as the market data behind it.
Twelve states are running the pilot: California, Colorado, Connecticut, Florida, Iowa, Louisiana, Maryland, Pennsylvania, Rhode Island, Vermont, Virginia, and Wisconsin. Neither Carolina is among them. That does not mean the Carolinas are outside the conversation, but it does mean the two states arrive at this from very different starting points.
North Carolina and South Carolina Are Not in the Same Position
This is where the two states diverge, and the gap is wider than most people would guess.
North Carolina
North Carolina adopted the NAIC Model Bulletin on the use of AI systems by insurers in December 2024, through Bulletin 24-B-19, without material changes to the model language. That bulletin does several things that matter to a claimant. It states that existing insurance law applies to AI-driven decisions exactly as it applies to decisions made by a human adjuster. It expects insurers to maintain a written AI governance program. And it holds the insurer responsible for AI systems acquired from third-party vendors rather than letting them point at the software company.
South Carolina
South Carolina has taken no formal state-level action on insurer AI use as of mid-2026. It has not adopted the Model Bulletin and it does not operate under an alternative framework the way California, Colorado, New York, and Texas do. South Carolina drivers are not without protection, because general unfair claim settlement practice law still applies, but there is no AI-specific standard to point to.
| Question | North Carolina | South Carolina |
|---|---|---|
| NAIC AI Model Bulletin adopted | Yes, Bulletin 24-B-19 (Dec 2024) | No formal action to date |
| Written AI governance program expected | Yes, under the bulletin | Not required by state guidance |
| Insurer accountable for vendor AI | Stated explicitly | No AI-specific state standard |
| In the NAIC evaluation tool pilot | No | No |
| Practical leverage on your claim | A documented standard you can cite | General unfair practice law only |
If you are filing in North Carolina, you have a written state standard to reference when you ask how a number was produced. If you are filing in South Carolina, you are relying on general claim handling law, which is real but less specific. In both states the practical answer is the same: bring your own evidence.
Why the Software Said So Is Not an Answer
The core principle behind the NAIC bulletin is simple and worth committing to memory. Existing insurance law applies to an AI-driven decision the same way it applies to a decision made by a person. Automating a determination does not create a new category of decision that escapes the rules.
The second principle is just as useful. When an insurer licenses a valuation or damage-assessment platform from an outside vendor, responsibility for the output stays with the insurer. An adjuster who tells you the number came from the system and there is nothing they can do is describing an internal process, not a legal limit.
An automated valuation is a position, not a verdict. It is the insurer’s opening number, produced by a tool they selected, using data they chose. You are entitled to understand the basis for it and to put competing evidence in front of them.
How This Shows Up in a Carolina Diminished Value Claim
The failure modes are consistent, and they are not exotic. Three come up repeatedly:
Severity Read From Photographs Alone
Image analysis reads what is visible in the frame. Structural repair, weld work, and sensor recalibration are frequently invisible in a post-repair photo. A model working from images can classify a repair as moderate when the shop’s line-by-line order describes something much more serious. Severity drives the diminished value figure, so an understated severity read produces an understated claim.
Comparable Data That Does Not Match the Local Market
A model is only as good as the market data behind it. Comparables pulled from outside Charlotte, Raleigh, Greenville, or Charleston can carry pricing that does not reflect what a buyer in your market would actually pay. That is a data quality question, and it is exactly the category the NAIC framework asks insurers about.
Fault and Severity Scored Together
This one is specific to North Carolina. Because the state applies contributory negligence, any assigned fault on your part can bar recovery entirely. When an automated system contributes to an early liability read, that read becomes the frame the entire claim is built on. Our guide on diminished value claims in North Carolina covers how much weight the fault determination carries there.
What to Ask and How to Ask It
- Ask whether an automated system produced the valuation. Put the question in writing. A file note that you asked is worth more than a phone conversation nobody recorded.
- Request the underlying comparables and adjustments. Ask for the full valuation document, not the summary letter, including every comparable vehicle and every adjustment applied.
- Ask which market the comparables came from. Distance from your market is one of the most common and most correctable errors in a valuation.
- Point to the repair order, not the photos. The shop’s line-by-line documentation describes work that image analysis cannot see. That document is your strongest correction to an understated severity read.
- In North Carolina, reference the state bulletin. Bulletin 24-B-19 establishes that AI-driven decisions are held to the same standards as human ones and that the insurer owns the output of vendor systems.
- Escalate if you get nowhere. Both the North Carolina Department of Insurance and the South Carolina Department of Insurance accept consumer complaints on claim handling. The NAIC’s public material on insurer AI use is useful background before you file one.
The Appraisal Is Still the Counterweight
Regulatory frameworks move slowly. The evaluation tool is not expected to be adopted until November 2026, and even then it governs how examiners review insurers rather than how any individual claim gets decided. None of it helps the driver whose repaired vehicle is sitting in a driveway right now with an offer on the table.
What does help is putting a competing number in front of the adjuster. An independent appraisal is a human inspection, a stated methodology, and a professional signature. It is evidence a model cannot dismiss and an adjuster has to respond to on the merits. Our breakdown of how to prove and maximize a diminished value claim covers how to present it, and the diminished value calculator gives you a working baseline before you start the conversation.
Get a Human Valuation of Your Carolina Claim
No algorithm, no formula shortcut. A real inspection and a defensible number you can put in front of the adjuster. Get Your Free Quote
Download This Guide as a PDF
Includes the NC versus SC comparison table and the list of questions to put to your adjuster. Download the PDF
Frequently Asked Questions
Is it legal for my insurer to use AI to decide my claim?
Yes. What regulators have established is that using AI does not change the legal standards that apply. Existing insurance law governs an AI-driven decision the same way it governs one made by a person, and the insurer remains responsible for the output even when the software came from an outside vendor.
How do I find out if AI was used on my diminished value claim?
Ask, in writing. Insurers are not generally required to volunteer which internal tools produced a number, but asking creates a record and it frequently prompts a more detailed explanation than a verbal exchange would. Pair the question with a request for the full valuation document and the comparables behind it.
Does North Carolina regulate insurer AI differently than South Carolina?
Yes. North Carolina adopted the NAIC Model Bulletin in December 2024 through Bulletin 24-B-19, which sets expectations for AI governance and confirms that insurers are accountable for third-party AI systems. South Carolina has not adopted the bulletin or an equivalent framework, so claimants there rely on general unfair claim settlement practice law.
Can I dispute a valuation that was generated by software?
Yes, and the process is the same as disputing any other valuation. Request the full report, check the comparables and adjustments, correct any understated damage severity using the repair order, and submit an independent appraisal with a competing figure. The origin of the number does not change your right to challenge it.
Will the NAIC evaluation tool help my current claim?
Not directly. The tool is a framework for state examiners to review how insurers govern their AI systems, not a mechanism for individual claim appeals. It is expected to be adopted in November 2026. For a claim on the table today, documentation and an independent appraisal are what move the number.
Why would an automated system understate a diminished value claim?
Usually because of what it cannot see. Image analysis reads visible damage, while structural repair, welding, and sensor recalibration often leave no visual trace after paint. If severity is scored from photographs, the model can classify a serious repair as moderate, and severity is one of the main drivers of the final figure.
