AI belongs on the parts of a property damage claim that are reading and sorting work: pulling the relevant clauses out of a 90-page policy, putting 400 photos in order, building a first-pass timeline out of the correspondence, drafting a letter someone then edits. It does not belong anywhere a licensed public adjuster's loyalty, judgment, or signature is the thing being relied on. That line is not a matter of taste. It follows from what a public adjuster license already obligates you to do.

Here is where the line sits, what the carrier on the other side is already running, and which rules apply to whom.

What this guide covers

  • How much AI insurers already use on property claims
  • Whether AI rules apply to public adjusters or only to insurers
  • The work AI does well on a claim file
  • The work that has to stay with a licensed human
  • Why an AI mistake costs more inside a claim file
  • Guardrails worth writing down

How much AI is already on the other side of your claims?

More than most policyholders assume. The NAIC has surveyed insurers by line of business since 2021. Of the 194 home insurers that responded, 70 percent said they use, plan to use, or plan to explore AI or machine learning in their operations. For auto insurers the figure was 88 percent of 193 companies.

On the property and casualty side, the NAIC reports that insurers use AI in claims for "accident image analysis and to estimate ultimate claim settlement values, along with fraud detection." Read that plainly: on a lot of files, something has already scored the photos and produced a settlement range before a human formed an opinion about the loss. Our post on whether AI can adjust insurance claims goes through what those systems do and do not decide.

Do the AI rules apply to public adjusters, or only to insurers?

Mostly to insurers, which surprises people. The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted in December 2023, is addressed to "All Insurers Licensed to Do Business" in the adopting state. It expects insurers to run a written AI program covering governance, risk controls, and internal audit, and it tells them what a regulator may ask for during an examination. Per the NAIC's own implementation map, 24 states and the District of Columbia had adopted the bulletin as of April 1, 2026, Maryland and DC among them. California, Colorado, New York, and Texas have their own insurance-specific AI guidance instead.

None of that governs a public adjusting firm directly. What governs you is your state's public adjuster licensing law, and those duties are older and blunter than any AI guidance. That is the more useful way to think about it. AI does not hand a public adjuster a new obligation so much as a new way to breach an old one.

What does AI do well on a property claim?

Four things, all of them reading work.

Policy retrieval. Finding the water damage exclusion, the ordinance and law limit, the appraisal clause, and the suit limitation inside a long form, with page references, is fast and easy to verify.

Document triage. Sorting a production of photos, invoices, and letters by date, room, and coverage part removes hours of clerical work from a large loss.

Timeline building. Pulling every date out of the correspondence gives you a first draft of the chronology, which matters because deadlines decide claims more often than arguments do.

Drafting. A first pass at a coverage position or a supplement narrative is a starting point, and starting points are cheap.

The pattern behind all four is the same. AI earns its place when a human is going to check the output against a source document anyway.

What has to stay with a licensed human?

Three duties, taken straight from the licensing rules. The NAIC Public Adjuster Licensing Model Act says a public adjuster is "obligated, under his or her license, to serve with objectivity and complete loyalty the interest of his client alone." Loyalty is not a function you can call. It is the reason the license exists at all.

The same act says a public adjuster "shall not undertake the adjustment of any claim if the public adjuster is not competent and knowledgeable as to the terms and conditions of the insurance coverage." Software that reads a policy does not transfer competence to the person holding the software. If you cannot explain why the coverage read is right, you are not in a position to sign your name under it.

And a public adjuster "may not agree to any loss settlement without the insured's knowledge and consent." So nothing automated settles a claim. A tool can prepare the number. A person presents it, and the insured decides. Our guide on the duties of a public adjuster covers the rest of the role.

State versions of these provisions differ, so read your own statute. The direction is the same everywhere: the license attaches to a human who can be questioned, disciplined, and held to a duty.

Why does an AI mistake cost more inside a claim file?

Because claim files are permanent, and they get read by people looking for a problem. Language models produce fluent text that is sometimes wrong, and the NAIC says so in plain terms: these tools "do not truly understand context and meaning the way humans do and may generate information that sounds accurate but is incorrect," so AI-generated information "should be reviewed carefully, especially when used for important decisions."

Now put that inside a public adjuster's file. An invented policy provision in a coverage letter hands the carrier a credibility argument it did not have to earn. A wrong square footage in a sworn statement in proof of loss is a sworn wrong number. And under the model act's record retention section, the file stays "for at least five (5) years after the termination of the transaction with an insured" and is "open to examination by the commissioner at all times." An error nobody caught does not quietly go away when the claim closes.

What guardrails should a firm write down?

Keep the list short enough that people actually follow it.

  • Check every AI output against the source document before it leaves the firm, and cite the page.
  • Name one licensed adjuster who reviews and owns each file, whether or not AI touched it.
  • Keep client information out of any tool that trains on your inputs, and confirm that in the contract rather than on the marketing page.
  • Let nothing automated send a settlement communication or accept an offer.
  • Log which tool produced what, since the output becomes part of a record you keep for years.
  • Tell clients plainly that you use software to analyze claims and that a licensed adjuster owns the file.

Where does Clayem fit?

Clayem is the leading AI claims-analysis platform for the recovery side, built for public adjusting firms and policyholders rather than carriers. It reads the policy, the photos, and the carrier's letters, returns a coverage read with page citations, counts down the deadlines, and shows the next step. The citations are the design choice that matters here, because an analysis you can check against a page number is one you can defend. It supports the licensed adjuster's judgment rather than standing in for it. For the operational half of the stack, see our guide to choosing a public adjuster CRM.

The bottom line

Use AI for reading, sorting, and drafting. Keep loyalty, judgment, and consent with the licensed human whose name is on the contract. Firms that get this backwards rarely fail loudly. They fail in one file, in one sentence nobody checked.

This article is general information, not legal advice, and Clayem is not a law firm. Public adjuster licensing rules and state AI guidance differ, so confirm your obligations with your state insurance department or an attorney licensed in your state.